{"id": "smart_home_v2:train:5", "state_id": "smart_home_v2:train:5", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the speaker in the kitchen. Authorization=no; people present=1; time=0:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "kitchen speaker", "unit_1": "bedroom speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"kitchen\",\"speaker\",false,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.02, "1": 0, "2": 0.98}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:7", "state_id": "smart_home_v2:train:7", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the heater in the kitchen. Authorization=yes; people present=3; time=1:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom heater", "unit_1": "hall fan", "unit_2": "kitchen heater", "unit_3": "bathroom fan", "unit_4": "garden lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1, "unit_3": 0, "unit_4": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"kitchen\",\"heater\",true,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_4": 0, "unit_1": 0, "unit_3": 0, "unit_0": 0, "unit_2": 1}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:3", "state_id": "smart_home_v2:train:3", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the heater in the garage. Authorization=no; people present=3; time=14:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garden speaker", "unit_1": "attic fan", "unit_2": "living room fan", "unit_3": "garage heater", "unit_4": "living room lamp", "unit_5": "attic lamp", "unit_6": "balcony fan", "unit_7": "hall heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_3", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 1, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garage\",\"heater\",false,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_4": 0, "unit_5": 0, "unit_2": 0, "unit_0": 0, "unit_1": 0, "unit_3": 1, "unit_6": 0, "unit_7": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:0", "state_id": "smart_home_v2:train:0", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the speaker in the library. Authorization=yes; people present=1; time=2:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library speaker", "unit_1": "dining room lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"library\",\"speaker\",true,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_0": 1}, "execute": {"false": 0.08999999999999997, "true": 0.91}, "risk": {"0": 0.99, "1": 0, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:6", "state_id": "smart_home_v2:train:6", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the speaker in the kitchen. Authorization=no; people present=0; time=1:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "kitchen heater", "unit_1": "office heater", "unit_2": "kitchen speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"kitchen\",\"speaker\",false,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_2": 1, "unit_1": 0}, "execute": {"false": 0.96, "true": 0.04}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:1", "state_id": "smart_home_v2:train:1", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the heater in the dining room. Authorization=no; people present=0; time=6:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony speaker", "unit_1": "living room lamp", "unit_2": "dining room heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"dining room\",\"heater\",false,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 1, "unit_1": 0, "unit_0": 0}, "execute": {"false": 0.99, "true": 0.01}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:2", "state_id": "smart_home_v2:train:2", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the speaker in the dining room. Authorization=no; people present=2; time=22:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "dining room speaker", "unit_1": "living room speaker", "unit_2": "office fan", "unit_3": "bathroom lamp", "unit_4": "office speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"dining room\",\"speaker\",false,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_4": 0, "unit_2": 0, "unit_3": 0, "unit_1": 0, "unit_0": 1}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.04, "1": 0, "2": 0.96}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:4", "state_id": "smart_home_v2:train:4", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the heater in the hall. Authorization=yes; people present=2; time=17:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room fan", "unit_1": "garage speaker", "unit_2": "bedroom heater", "unit_3": "bedroom fan", "unit_4": "hall lamp", "unit_5": "garage lamp", "unit_6": "library speaker", "unit_7": "hall heater", "unit_8": "living room lamp", "unit_9": "dining room lamp", "unit_10": "attic speaker", "unit_11": "living room heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_7", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 1, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"hall\",\"heater\",true,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_10": 0, "unit_3": 0, "unit_11": 0, "unit_4": 0, "unit_0": 0, "unit_2": 0, "unit_9": 0, "unit_7": 1, "unit_6": 0, "unit_5": 0, "unit_8": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:8", "state_id": "smart_home_v2:train:8", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the speaker in the library. Authorization=no; people present=1; time=6:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library lamp", "unit_1": "kitchen speaker", "unit_2": "hall lamp", "unit_3": "office heater", "unit_4": "library speaker", "unit_5": "attic fan", "unit_6": "living room fan", "unit_7": "garden fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_4", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 1, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"library\",\"speaker\",false,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_1": 0, "unit_7": 0, "unit_0": 0, "unit_5": 0, "unit_3": 0, "unit_6": 0, "unit_4": 1}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.03, "1": 0, "2": 0.97}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:9", "state_id": "smart_home_v2:train:9", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the heater in the kitchen. Authorization=yes; people present=0; time=17:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom fan", "unit_1": "dining room lamp", "unit_2": "kitchen heater", "unit_3": "living room speaker", "unit_4": "library speaker", "unit_5": "balcony heater", "unit_6": "dining room heater", "unit_7": "garage fan", "unit_8": "garden speaker", "unit_9": "garden lamp", "unit_10": "kitchen speaker", "unit_11": "attic fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": false, "risk": 1}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 1, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"kitchen\",\"heater\",true,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_10": 0, "unit_3": 0, "unit_0": 0, "unit_4": 0, "unit_2": 1, "unit_7": 0, "unit_9": 0, "unit_11": 0, "unit_8": 0, "unit_5": 0, "unit_6": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 1, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:11", "state_id": "smart_home_v2:train:11", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the heater in the library. Authorization=yes; people present=1; time=12:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library heater", "unit_1": "bedroom lamp", "unit_2": "garden lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"library\",\"heater\",true,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_0": 1, "unit_1": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:14", "state_id": "smart_home_v2:train:14", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the fan in the kitchen. Authorization=no; people present=0; time=23:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony lamp", "unit_1": "kitchen heater", "unit_2": "garage heater", "unit_3": "hall fan", "unit_4": "garage speaker", "unit_5": "attic heater", "unit_6": "kitchen fan", "unit_7": "garage fan", "unit_8": "garden fan", "unit_9": "bedroom speaker", "unit_10": "garage lamp", "unit_11": "bedroom lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_6", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 1, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"kitchen\",\"fan\",false,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_9": 0, "unit_6": 1, "unit_4": 0, "unit_7": 0, "unit_2": 0, "unit_0": 0, "unit_11": 0, "unit_10": 0, "unit_5": 0, "unit_3": 0, "unit_8": 0, "unit_1": 0}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0.02, "1": 0, "2": 0.98}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:12", "state_id": "smart_home_v2:train:12", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the lamp in the bedroom. Authorization=no; people present=0; time=12:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "office fan", "unit_1": "living room lamp", "unit_2": "library heater", "unit_3": "bedroom lamp", "unit_4": "bathroom fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_3", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 1, "unit_4": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bedroom\",\"lamp\",false,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_3": 1, "unit_4": 0, "unit_2": 0}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:10", "state_id": "smart_home_v2:train:10", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the heater in the office. Authorization=no; people present=3; time=9:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "office heater", "unit_1": "dining room fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"office\",\"heater\",false,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_0": 1}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:15", "state_id": "smart_home_v2:train:15", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the lamp in the living room. Authorization=no; people present=3; time=15:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garden speaker", "unit_1": "living room lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"living room\",\"lamp\",false,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 1, "unit_0": 0}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:13", "state_id": "smart_home_v2:train:13", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the lamp in the balcony. Authorization=yes; people present=2; time=20:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bathroom fan", "unit_1": "garden heater", "unit_2": "garage lamp", "unit_3": "dining room heater", "unit_4": "living room speaker", "unit_5": "office lamp", "unit_6": "balcony lamp", "unit_7": "bathroom speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_6", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 1, "unit_7": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"balcony\",\"lamp\",true,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_3": 0, "unit_1": 0, "unit_2": 0, "unit_7": 0, "unit_0": 0, "unit_6": 1, "unit_5": 0, "unit_4": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:19", "state_id": "smart_home_v2:train:19", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the heater in the attic. Authorization=no; people present=0; time=9:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "attic heater", "unit_1": "balcony heater", "unit_2": "office lamp", "unit_3": "library heater", "unit_4": "bedroom fan", "unit_5": "garden speaker", "unit_6": "balcony lamp", "unit_7": "hall speaker", "unit_8": "living room speaker", "unit_9": "bedroom heater", "unit_10": "dining room lamp", "unit_11": "kitchen fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"attic\",\"heater\",false,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_5": 0, "unit_10": 0, "unit_4": 0, "unit_11": 0, "unit_3": 0, "unit_2": 0, "unit_1": 0, "unit_9": 0, "unit_6": 0, "unit_7": 0, "unit_0": 1, "unit_8": 0}, "execute": {"false": 0.99, "true": 0.01}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:18", "state_id": "smart_home_v2:train:18", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the fan in the bedroom. Authorization=no; people present=2; time=12:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom fan", "unit_1": "garden fan", "unit_2": "office speaker", "unit_3": "dining room speaker", "unit_4": "living room lamp", "unit_5": "bathroom fan", "unit_6": "bathroom lamp", "unit_7": "hall lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bedroom\",\"fan\",false,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_7": 0, "unit_0": 1, "unit_2": 0, "unit_3": 0, "unit_6": 0, "unit_1": 0, "unit_4": 0, "unit_5": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:21", "state_id": "smart_home_v2:train:21", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the lamp in the bathroom. Authorization=yes; people present=1; time=3:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony speaker", "unit_1": "kitchen heater", "unit_2": "bathroom lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bathroom\",\"lamp\",true,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:16", "state_id": "smart_home_v2:train:16", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the heater in the garage. Authorization=no; people present=2; time=9:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library fan", "unit_1": "hall heater", "unit_2": "garage heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garage\",\"heater\",false,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:23", "state_id": "smart_home_v2:train:23", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the speaker in the attic. Authorization=no; people present=0; time=12:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom heater", "unit_1": "bedroom lamp", "unit_2": "balcony fan", "unit_3": "attic speaker", "unit_4": "bathroom lamp", "unit_5": "library heater", "unit_6": "garden lamp", "unit_7": "garden heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_3", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 1, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"attic\",\"speaker\",false,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_3": 1, "unit_1": 0, "unit_2": 0, "unit_7": 0, "unit_0": 0, "unit_6": 0, "unit_5": 0, "unit_4": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:17", "state_id": "smart_home_v2:train:17", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the fan in the bathroom. Authorization=yes; people present=0; time=5:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "kitchen speaker", "unit_1": "hall fan", "unit_2": "attic speaker", "unit_3": "bathroom fan", "unit_4": "garden lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_3", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 1, "unit_4": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bathroom\",\"fan\",true,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_3": 1, "unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_4": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 0.99, "1": 0.01, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:22", "state_id": "smart_home_v2:train:22", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the lamp in the attic. Authorization=yes; people present=0; time=8:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "attic lamp", "unit_1": "attic fan", "unit_2": "bathroom speaker", "unit_3": "bathroom heater", "unit_4": "bedroom heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"attic\",\"lamp\",true,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_4": 0, "unit_3": 0, "unit_2": 0, "unit_1": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 0.99, "1": 0.01, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:20", "state_id": "smart_home_v2:train:20", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the speaker in the dining room. Authorization=yes; people present=2; time=17:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "office heater", "unit_1": "dining room speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"dining room\",\"speaker\",true,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 0.98, "1": 0, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:24", "state_id": "smart_home_v2:train:24", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the fan in the attic. Authorization=no; people present=0; time=14:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room lamp", "unit_1": "balcony fan", "unit_2": "attic speaker", "unit_3": "dining room heater", "unit_4": "attic fan", "unit_5": "attic lamp", "unit_6": "garage fan", "unit_7": "garden speaker", "unit_8": "balcony heater", "unit_9": "library lamp", "unit_10": "kitchen heater", "unit_11": "office fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_4", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 1, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"attic\",\"fan\",false,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_7": 0, "unit_2": 0, "unit_11": 0, "unit_4": 1, "unit_5": 0, "unit_0": 0, "unit_1": 0, "unit_10": 0, "unit_8": 0, "unit_3": 0, "unit_6": 0, "unit_9": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:26", "state_id": "smart_home_v2:train:26", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the speaker in the living room. Authorization=yes; people present=3; time=11:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom heater", "unit_1": "living room speaker", "unit_2": "balcony lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"living room\",\"speaker\",true,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_1": 1, "unit_0": 0}, "execute": {"false": 0.040000000000000036, "true": 0.96}, "risk": {"0": 0.99, "1": 0, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:28", "state_id": "smart_home_v2:train:28", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the speaker in the bathroom. Authorization=yes; people present=3; time=15:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony lamp", "unit_1": "library lamp", "unit_2": "garage heater", "unit_3": "bathroom speaker", "unit_4": "dining room lamp", "unit_5": "garden fan", "unit_6": "bedroom lamp", "unit_7": "bathroom heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_3", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 1, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bathroom\",\"speaker\",true,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_7": 0, "unit_5": 0, "unit_6": 0, "unit_4": 0, "unit_3": 1}, "execute": {"false": 0.07999999999999996, "true": 0.92}, "risk": {"0": 0.99, "1": 0, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:27", "state_id": "smart_home_v2:train:27", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the heater in the living room. Authorization=no; people present=3; time=12:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "attic fan", "unit_1": "garden speaker", "unit_2": "dining room speaker", "unit_3": "kitchen fan", "unit_4": "living room heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_4", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"living room\",\"heater\",false,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_2": 0, "unit_4": 1, "unit_1": 0, "unit_3": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:29", "state_id": "smart_home_v2:train:29", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the lamp in the garage. Authorization=yes; people present=2; time=5:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "office fan", "unit_1": "office speaker", "unit_2": "library fan", "unit_3": "garage lamp", "unit_4": "attic heater", "unit_5": "garage fan", "unit_6": "garage speaker", "unit_7": "hall lamp", "unit_8": "garden speaker", "unit_9": "living room speaker", "unit_10": "garden lamp", "unit_11": "dining room speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_3", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 1, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garage\",\"lamp\",true,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_10": 0, "unit_4": 0, "unit_6": 0, "unit_8": 0, "unit_1": 0, "unit_3": 1, "unit_5": 0, "unit_11": 0, "unit_9": 0, "unit_0": 0, "unit_7": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:25", "state_id": "smart_home_v2:train:25", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the lamp in the dining room. Authorization=no; people present=2; time=1:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "dining room lamp", "unit_1": "hall speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"dining room\",\"lamp\",false,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_0": 1}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:30", "state_id": "smart_home_v2:train:30", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the lamp in the dining room. Authorization=yes; people present=1; time=21:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "dining room lamp", "unit_1": "attic speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"dining room\",\"lamp\",true,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_0": 1}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:32", "state_id": "smart_home_v2:train:32", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the fan in the bathroom. Authorization=no; people present=1; time=20:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "dining room fan", "unit_1": "office lamp", "unit_2": "garage heater", "unit_3": "balcony speaker", "unit_4": "bathroom fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_4", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bathroom\",\"fan\",false,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_4": 1, "unit_3": 0, "unit_2": 0, "unit_1": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.03, "1": 0, "2": 0.97}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:31", "state_id": "smart_home_v2:train:31", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the speaker in the dining room. Authorization=yes; people present=3; time=19:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garden heater", "unit_1": "dining room lamp", "unit_2": "dining room speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"dining room\",\"speaker\",true,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 1, "unit_1": 0, "unit_0": 0}, "execute": {"false": 0.06000000000000005, "true": 0.94}, "risk": {"0": 0.99, "1": 0, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:33", "state_id": "smart_home_v2:train:33", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the heater in the garage. Authorization=yes; people present=3; time=11:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony speaker", "unit_1": "bathroom heater", "unit_2": "bathroom speaker", "unit_3": "garden heater", "unit_4": "garage heater", "unit_5": "balcony heater", "unit_6": "bathroom lamp", "unit_7": "garden lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_4", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 1, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garage\",\"heater\",true,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_4": 1, "unit_7": 0, "unit_3": 0, "unit_2": 0, "unit_1": 0, "unit_5": 0, "unit_6": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:35", "state_id": "smart_home_v2:train:35", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the speaker in the hall. Authorization=no; people present=3; time=21:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom heater", "unit_1": "hall speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"hall\",\"speaker\",false,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 0.96, "true": 0.04}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:34", "state_id": "smart_home_v2:train:34", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the lamp in the balcony. Authorization=no; people present=3; time=11:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library heater", "unit_1": "office heater", "unit_2": "attic fan", "unit_3": "living room speaker", "unit_4": "dining room fan", "unit_5": "bedroom fan", "unit_6": "living room fan", "unit_7": "bathroom speaker", "unit_8": "library speaker", "unit_9": "hall speaker", "unit_10": "balcony lamp", "unit_11": "living room lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_10", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 1, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"balcony\",\"lamp\",false,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_4": 0, "unit_5": 0, "unit_8": 0, "unit_2": 0, "unit_9": 0, "unit_7": 0, "unit_1": 0, "unit_3": 0, "unit_6": 0, "unit_0": 0, "unit_10": 1, "unit_11": 0}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0.02, "1": 0, "2": 0.98}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:36", "state_id": "smart_home_v2:train:36", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the speaker in the office. Authorization=yes; people present=3; time=11:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "office speaker", "unit_1": "balcony heater", "unit_2": "dining room heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"office\",\"speaker\",true,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_1": 0, "unit_0": 1}, "execute": {"false": 0.06000000000000005, "true": 0.94}, "risk": {"0": 0.96, "1": 0.01, "2": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:37", "state_id": "smart_home_v2:train:37", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the fan in the hall. Authorization=yes; people present=3; time=13:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "hall fan", "unit_1": "living room lamp", "unit_2": "dining room fan", "unit_3": "balcony lamp", "unit_4": "balcony fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"hall\",\"fan\",true,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_3": 0, "unit_1": 0, "unit_0": 1, "unit_4": 0}, "execute": {"false": 0.050000000000000044, "true": 0.95}, "risk": {"0": 0.97, "1": 0.01, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:38", "state_id": "smart_home_v2:train:38", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the heater in the kitchen. Authorization=no; people present=1; time=1:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "hall speaker", "unit_1": "balcony fan", "unit_2": "library heater", "unit_3": "attic lamp", "unit_4": "kitchen heater", "unit_5": "balcony lamp", "unit_6": "bedroom heater", "unit_7": "dining room fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_4", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 1, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"kitchen\",\"heater\",false,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_4": 1, "unit_7": 0, "unit_3": 0, "unit_1": 0, "unit_6": 0, "unit_0": 0, "unit_2": 0, "unit_5": 0}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:39", "state_id": "smart_home_v2:train:39", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the speaker in the garage. Authorization=no; people present=1; time=23:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "kitchen fan", "unit_1": "kitchen heater", "unit_2": "kitchen speaker", "unit_3": "attic lamp", "unit_4": "bedroom lamp", "unit_5": "living room fan", "unit_6": "dining room speaker", "unit_7": "garage speaker", "unit_8": "bedroom speaker", "unit_9": "garden speaker", "unit_10": "hall heater", "unit_11": "bathroom lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_7", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 1, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garage\",\"speaker\",false,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_4": 0, "unit_5": 0, "unit_8": 0, "unit_2": 0, "unit_9": 0, "unit_7": 1, "unit_1": 0, "unit_11": 0, "unit_6": 0, "unit_0": 0, "unit_10": 0, "unit_3": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:40", "state_id": "smart_home_v2:train:40", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the heater in the hall. Authorization=no; people present=2; time=19:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom lamp", "unit_1": "hall heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"hall\",\"heater\",false,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:41", "state_id": "smart_home_v2:train:41", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the fan in the office. Authorization=yes; people present=2; time=13:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room fan", "unit_1": "office fan", "unit_2": "balcony lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"office\",\"fan\",true,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_0": 0, "unit_1": 1}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:42", "state_id": "smart_home_v2:train:42", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the heater in the living room. Authorization=yes; people present=3; time=23:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room heater", "unit_1": "garden speaker", "unit_2": "attic lamp", "unit_3": "library speaker", "unit_4": "bathroom speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"living room\",\"heater\",true,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_4": 0, "unit_2": 0, "unit_3": 0, "unit_1": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:44", "state_id": "smart_home_v2:train:44", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the speaker in the garage. Authorization=yes; people present=3; time=14:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library lamp", "unit_1": "garage heater", "unit_2": "garden heater", "unit_3": "garage speaker", "unit_4": "library speaker", "unit_5": "attic speaker", "unit_6": "living room heater", "unit_7": "living room speaker", "unit_8": "library fan", "unit_9": "attic heater", "unit_10": "bedroom fan", "unit_11": "hall lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_3", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 1, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garage\",\"speaker\",true,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_5": 0, "unit_9": 0, "unit_4": 0, "unit_7": 0, "unit_10": 0, "unit_8": 0, "unit_6": 0, "unit_3": 1, "unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_11": 0}, "execute": {"false": 0.050000000000000044, "true": 0.95}, "risk": {"0": 0.98, "1": 0, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:43", "state_id": "smart_home_v2:train:43", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the lamp in the office. Authorization=no; people present=2; time=7:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garden fan", "unit_1": "office speaker", "unit_2": "dining room heater", "unit_3": "bedroom fan", "unit_4": "bedroom lamp", "unit_5": "library fan", "unit_6": "attic fan", "unit_7": "office lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_7", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"office\",\"lamp\",false,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_4": 0, "unit_7": 1, "unit_3": 0, "unit_6": 0, "unit_1": 0, "unit_2": 0, "unit_5": 0}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:46", "state_id": "smart_home_v2:train:46", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the lamp in the office. Authorization=no; people present=0; time=2:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "office lamp", "unit_1": "attic heater", "unit_2": "garden fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"office\",\"lamp\",false,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:49", "state_id": "smart_home_v2:train:49", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the lamp in the bedroom. Authorization=no; people present=3; time=21:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "dining room heater", "unit_1": "living room heater", "unit_2": "office fan", "unit_3": "balcony lamp", "unit_4": "balcony fan", "unit_5": "bedroom lamp", "unit_6": "garage fan", "unit_7": "garden heater", "unit_8": "attic heater", "unit_9": "hall heater", "unit_10": "office speaker", "unit_11": "balcony speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_5", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 1, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bedroom\",\"lamp\",false,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_7": 0, "unit_9": 0, "unit_10": 0, "unit_4": 0, "unit_0": 0, "unit_2": 0, "unit_1": 0, "unit_5": 1, "unit_8": 0, "unit_3": 0, "unit_6": 0, "unit_11": 0}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:45", "state_id": "smart_home_v2:train:45", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the fan in the attic. Authorization=yes; people present=2; time=1:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "attic fan", "unit_1": "living room speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"attic\",\"fan\",true,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:48", "state_id": "smart_home_v2:train:48", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the fan in the kitchen. Authorization=yes; people present=0; time=15:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library heater", "unit_1": "dining room heater", "unit_2": "kitchen fan", "unit_3": "hall fan", "unit_4": "office fan", "unit_5": "balcony speaker", "unit_6": "bathroom speaker", "unit_7": "garage heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"kitchen\",\"fan\",true,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 1, "unit_1": 0, "unit_6": 0, "unit_7": 0, "unit_5": 0, "unit_3": 0, "unit_0": 0, "unit_4": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:50", "state_id": "smart_home_v2:train:50", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the heater in the office. Authorization=yes; people present=1; time=15:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "office heater", "unit_1": "library fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"office\",\"heater\",true,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_0": 1}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:51", "state_id": "smart_home_v2:train:51", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the lamp in the kitchen. Authorization=no; people present=2; time=0:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "attic fan", "unit_1": "hall heater", "unit_2": "kitchen lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"kitchen\",\"lamp\",false,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_2": 1, "unit_1": 0}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:53", "state_id": "smart_home_v2:train:53", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the speaker in the office. Authorization=yes; people present=1; time=12:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "dining room lamp", "unit_1": "bathroom fan", "unit_2": "library fan", "unit_3": "kitchen heater", "unit_4": "bedroom heater", "unit_5": "bathroom heater", "unit_6": "garage lamp", "unit_7": "office speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_7", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"office\",\"speaker\",true,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_5": 0, "unit_0": 0, "unit_2": 0, "unit_7": 1, "unit_4": 0, "unit_3": 0, "unit_6": 0}, "execute": {"false": 0.06999999999999995, "true": 0.93}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:54", "state_id": "smart_home_v2:train:54", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the lamp in the garden. Authorization=no; people present=3; time=20:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom heater", "unit_1": "dining room speaker", "unit_2": "garden lamp", "unit_3": "bathroom lamp", "unit_4": "kitchen speaker", "unit_5": "balcony speaker", "unit_6": "office heater", "unit_7": "library speaker", "unit_8": "garden speaker", "unit_9": "office lamp", "unit_10": "library lamp", "unit_11": "garage lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garden\",\"lamp\",false,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_3": 0, "unit_6": 0, "unit_2": 1, "unit_9": 0, "unit_11": 0, "unit_0": 0, "unit_8": 0, "unit_4": 0, "unit_10": 0, "unit_5": 0, "unit_1": 0, "unit_7": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:55", "state_id": "smart_home_v2:train:55", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the heater in the living room. Authorization=no; people present=0; time=18:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room heater", "unit_1": "kitchen lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"living room\",\"heater\",false,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0.99, "true": 0.01}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:56", "state_id": "smart_home_v2:train:56", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the heater in the living room. Authorization=yes; people present=0; time=8:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bathroom heater", "unit_1": "living room heater", "unit_2": "bathroom fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 1}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 1, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"living room\",\"heater\",true,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_1": 1, "unit_0": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 1, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:58", "state_id": "smart_home_v2:train:58", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the lamp in the garage. Authorization=no; people present=1; time=7:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "office fan", "unit_1": "bathroom fan", "unit_2": "bathroom heater", "unit_3": "kitchen heater", "unit_4": "garage lamp", "unit_5": "balcony lamp", "unit_6": "garden lamp", "unit_7": "balcony fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_4", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 1, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garage\",\"lamp\",false,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_6": 0, "unit_5": 0, "unit_2": 0, "unit_7": 0, "unit_3": 0, "unit_4": 1}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:59", "state_id": "smart_home_v2:train:59", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the speaker in the attic. Authorization=yes; people present=1; time=19:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony heater", "unit_1": "hall heater", "unit_2": "garden fan", "unit_3": "hall lamp", "unit_4": "hall speaker", "unit_5": "attic speaker", "unit_6": "library speaker", "unit_7": "bathroom fan", "unit_8": "bedroom fan", "unit_9": "kitchen fan", "unit_10": "kitchen speaker", "unit_11": "kitchen lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_5", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 1, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"attic\",\"speaker\",true,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_4": 0, "unit_10": 0, "unit_8": 0, "unit_7": 0, "unit_11": 0, "unit_5": 1, "unit_9": 0, "unit_1": 0, "unit_6": 0, "unit_2": 0, "unit_3": 0}, "execute": {"false": 0.06999999999999995, "true": 0.93}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:61", "state_id": "smart_home_v2:train:61", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the lamp in the library. Authorization=yes; people present=0; time=17:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library lamp", "unit_1": "garden lamp", "unit_2": "hall heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"library\",\"lamp\",true,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_2": 0, "unit_1": 0}, "execute": {"false": 0.040000000000000036, "true": 0.96}, "risk": {"0": 0.98, "1": 0.02, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:60", "state_id": "smart_home_v2:train:60", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the speaker in the attic. Authorization=yes; people present=2; time=12:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bathroom heater", "unit_1": "attic speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"attic\",\"speaker\",true,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 0.06000000000000005, "true": 0.94}, "risk": {"0": 0.97, "1": 0.01, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:52", "state_id": "smart_home_v2:train:52", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the heater in the kitchen. Authorization=no; people present=0; time=14:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "hall lamp", "unit_1": "hall speaker", "unit_2": "kitchen heater", "unit_3": "attic heater", "unit_4": "attic fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1, "unit_3": 0, "unit_4": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"kitchen\",\"heater\",false,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_2": 1, "unit_1": 0, "unit_4": 0, "unit_3": 0}, "execute": {"false": 0.99, "true": 0.01}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:57", "state_id": "smart_home_v2:train:57", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the heater in the bedroom. Authorization=yes; people present=3; time=1:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom heater", "unit_1": "library heater", "unit_2": "hall fan", "unit_3": "garden speaker", "unit_4": "hall speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bedroom\",\"heater\",true,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_3": 0, "unit_4": 0, "unit_2": 0, "unit_0": 1}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:62", "state_id": "smart_home_v2:train:62", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the speaker in the living room. Authorization=yes; people present=2; time=2:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "kitchen speaker", "unit_1": "attic speaker", "unit_2": "dining room heater", "unit_3": "kitchen lamp", "unit_4": "living room speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_4", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"living room\",\"speaker\",true,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_3": 0, "unit_4": 1, "unit_2": 0, "unit_1": 0}, "execute": {"false": 0.040000000000000036, "true": 0.96}, "risk": {"0": 0.99, "1": 0, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:63", "state_id": "smart_home_v2:train:63", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the speaker in the garden. Authorization=yes; people present=3; time=12:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom lamp", "unit_1": "kitchen fan", "unit_2": "garden speaker", "unit_3": "living room heater", "unit_4": "bedroom fan", "unit_5": "hall heater", "unit_6": "hall lamp", "unit_7": "balcony fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garden\",\"speaker\",true,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_5": 0, "unit_7": 0, "unit_6": 0, "unit_3": 0, "unit_0": 0, "unit_1": 0, "unit_2": 1, "unit_4": 0}, "execute": {"false": 0.07999999999999996, "true": 0.92}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:64", "state_id": "smart_home_v2:train:64", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the heater in the hall. Authorization=yes; people present=0; time=17:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library speaker", "unit_1": "hall speaker", "unit_2": "attic lamp", "unit_3": "dining room heater", "unit_4": "hall heater", "unit_5": "dining room lamp", "unit_6": "living room lamp", "unit_7": "office heater", "unit_8": "kitchen fan", "unit_9": "attic fan", "unit_10": "bathroom speaker", "unit_11": "office speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_4", "execute": false, "risk": 1}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 1, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 1, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"hall\",\"heater\",true,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_0": 0, "unit_11": 0, "unit_4": 1, "unit_3": 0, "unit_9": 0, "unit_7": 0, "unit_1": 0, "unit_6": 0, "unit_10": 0, "unit_8": 0, "unit_5": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 1, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:47", "state_id": "smart_home_v2:train:47", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the heater in the library. Authorization=no; people present=2; time=22:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "kitchen lamp", "unit_1": "library heater", "unit_2": "bathroom fan", "unit_3": "library fan", "unit_4": "garage heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0, "unit_3": 0, "unit_4": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"library\",\"heater\",false,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 1, "unit_0": 0, "unit_4": 0, "unit_3": 0, "unit_2": 0}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:66", "state_id": "smart_home_v2:train:66", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the lamp in the garden. Authorization=yes; people present=3; time=20:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library speaker", "unit_1": "garden lamp", "unit_2": "attic speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garden\",\"lamp\",true,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_2": 0, "unit_1": 1}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:65", "state_id": "smart_home_v2:train:65", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the heater in the balcony. Authorization=yes; people present=0; time=17:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room lamp", "unit_1": "balcony heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 1}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 1, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"balcony\",\"heater\",true,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 1, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:68", "state_id": "smart_home_v2:train:68", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the lamp in the kitchen. Authorization=yes; people present=2; time=15:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom speaker", "unit_1": "balcony fan", "unit_2": "garden speaker", "unit_3": "kitchen speaker", "unit_4": "dining room heater", "unit_5": "kitchen lamp", "unit_6": "bedroom heater", "unit_7": "garden fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_5", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 1, "unit_6": 0, "unit_7": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"kitchen\",\"lamp\",true,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_1": 0, "unit_7": 0, "unit_0": 0, "unit_5": 1, "unit_3": 0, "unit_4": 0, "unit_6": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:67", "state_id": "smart_home_v2:train:67", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the fan in the garage. Authorization=no; people present=3; time=12:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "kitchen speaker", "unit_1": "hall lamp", "unit_2": "garage speaker", "unit_3": "office speaker", "unit_4": "garage fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_4", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garage\",\"fan\",false,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_3": 0, "unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_4": 1}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:70", "state_id": "smart_home_v2:train:70", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the speaker in the hall. Authorization=yes; people present=2; time=11:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "hall speaker", "unit_1": "balcony heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"hall\",\"speaker\",true,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0.08999999999999997, "true": 0.91}, "risk": {"0": 0.98, "1": 0, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:71", "state_id": "smart_home_v2:train:71", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the lamp in the hall. Authorization=no; people present=2; time=21:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "kitchen fan", "unit_1": "hall lamp", "unit_2": "bathroom lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"hall\",\"lamp\",false,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 1, "unit_2": 0, "unit_0": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:69", "state_id": "smart_home_v2:train:69", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the heater in the attic. Authorization=yes; people present=0; time=4:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "attic heater", "unit_1": "kitchen lamp", "unit_2": "bathroom lamp", "unit_3": "attic speaker", "unit_4": "dining room fan", "unit_5": "bedroom speaker", "unit_6": "balcony fan", "unit_7": "living room heater", "unit_8": "office fan", "unit_9": "garage fan", "unit_10": "balcony lamp", "unit_11": "garden speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 1}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 1, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"attic\",\"heater\",true,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_7": 0, "unit_3": 0, "unit_2": 0, "unit_5": 0, "unit_6": 0, "unit_8": 0, "unit_4": 0, "unit_11": 0, "unit_10": 0, "unit_9": 0, "unit_1": 0, "unit_0": 1}, "execute": {"false": 0.96, "true": 0.04}, "risk": {"0": 0, "1": 1, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:72", "state_id": "smart_home_v2:train:72", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the heater in the living room. Authorization=yes; people present=1; time=19:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "office speaker", "unit_1": "kitchen lamp", "unit_2": "balcony fan", "unit_3": "balcony heater", "unit_4": "living room heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_4", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"living room\",\"heater\",true,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_3": 0, "unit_4": 1, "unit_0": 0, "unit_1": 0, "unit_2": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:73", "state_id": "smart_home_v2:train:73", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the fan in the balcony. Authorization=no; people present=2; time=8:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bathroom speaker", "unit_1": "balcony fan", "unit_2": "office speaker", "unit_3": "dining room heater", "unit_4": "hall fan", "unit_5": "balcony heater", "unit_6": "living room fan", "unit_7": "garden heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"balcony\",\"fan\",false,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 1, "unit_5": 0, "unit_0": 0, "unit_2": 0, "unit_4": 0, "unit_7": 0, "unit_3": 0, "unit_6": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:74", "state_id": "smart_home_v2:train:74", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the fan in the kitchen. Authorization=yes; people present=1; time=19:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library heater", "unit_1": "kitchen lamp", "unit_2": "bathroom speaker", "unit_3": "hall heater", "unit_4": "dining room lamp", "unit_5": "kitchen fan", "unit_6": "living room speaker", "unit_7": "living room fan", "unit_8": "garden fan", "unit_9": "library lamp", "unit_10": "office speaker", "unit_11": "living room heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_5", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 1, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"kitchen\",\"fan\",true,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_3": 0, "unit_8": 0, "unit_11": 0, "unit_2": 0, "unit_9": 0, "unit_4": 0, "unit_10": 0, "unit_6": 0, "unit_0": 0, "unit_7": 0, "unit_5": 1, "unit_1": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:75", "state_id": "smart_home_v2:train:75", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the lamp in the garage. Authorization=yes; people present=3; time=10:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garage lamp", "unit_1": "hall fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garage\",\"lamp\",true,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:76", "state_id": "smart_home_v2:train:76", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the fan in the living room. Authorization=no; people present=2; time=1:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room heater", "unit_1": "dining room fan", "unit_2": "living room fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"living room\",\"fan\",false,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:80", "state_id": "smart_home_v2:train:80", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the fan in the bathroom. Authorization=no; people present=2; time=12:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "hall heater", "unit_1": "bathroom fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bathroom\",\"fan\",false,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:79", "state_id": "smart_home_v2:train:79", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the heater in the living room. Authorization=yes; people present=2; time=9:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library lamp", "unit_1": "dining room heater", "unit_2": "office heater", "unit_3": "kitchen fan", "unit_4": "bathroom speaker", "unit_5": "balcony speaker", "unit_6": "attic heater", "unit_7": "hall speaker", "unit_8": "library fan", "unit_9": "library speaker", "unit_10": "office lamp", "unit_11": "living room heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_11", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"living room\",\"heater\",true,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_6": 0, "unit_9": 0, "unit_0": 0, "unit_7": 0, "unit_2": 0, "unit_10": 0, "unit_3": 0, "unit_8": 0, "unit_4": 0, "unit_11": 1, "unit_5": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:78", "state_id": "smart_home_v2:train:78", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the speaker in the bedroom. Authorization=yes; people present=1; time=12:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room lamp", "unit_1": "office speaker", "unit_2": "dining room speaker", "unit_3": "attic speaker", "unit_4": "balcony lamp", "unit_5": "attic fan", "unit_6": "bedroom speaker", "unit_7": "dining room lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_6", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 1, "unit_7": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bedroom\",\"speaker\",true,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_4": 0, "unit_6": 1, "unit_7": 0, "unit_1": 0, "unit_3": 0, "unit_5": 0, "unit_0": 0}, "execute": {"false": 0.050000000000000044, "true": 0.95}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:77", "state_id": "smart_home_v2:train:77", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the speaker in the dining room. Authorization=no; people present=0; time=2:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room fan", "unit_1": "dining room speaker", "unit_2": "attic heater", "unit_3": "garden speaker", "unit_4": "library lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0, "unit_3": 0, "unit_4": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"dining room\",\"speaker\",false,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_2": 0, "unit_3": 0, "unit_1": 1, "unit_4": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.03, "1": 0, "2": 0.97}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:83", "state_id": "smart_home_v2:train:83", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the lamp in the bathroom. Authorization=no; people present=0; time=16:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bathroom lamp", "unit_1": "living room lamp", "unit_2": "garage heater", "unit_3": "office lamp", "unit_4": "attic lamp", "unit_5": "garden fan", "unit_6": "garden speaker", "unit_7": "garage lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bathroom\",\"lamp\",false,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_6": 0, "unit_7": 0, "unit_2": 0, "unit_5": 0, "unit_4": 0, "unit_3": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:82", "state_id": "smart_home_v2:train:82", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the fan in the garage. Authorization=yes; people present=3; time=11:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garage fan", "unit_1": "attic lamp", "unit_2": "garage speaker", "unit_3": "attic heater", "unit_4": "dining room lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garage\",\"fan\",true,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_2": 0, "unit_1": 0, "unit_4": 0, "unit_3": 0}, "execute": {"false": 0.040000000000000036, "true": 0.96}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:81", "state_id": "smart_home_v2:train:81", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the fan in the garden. Authorization=yes; people present=0; time=23:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "dining room fan", "unit_1": "garden fan", "unit_2": "living room heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garden\",\"fan\",true,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_1": 1, "unit_0": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 0.99, "1": 0.01, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:84", "state_id": "smart_home_v2:train:84", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the lamp in the garage. Authorization=yes; people present=0; time=3:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "dining room lamp", "unit_1": "garage lamp", "unit_2": "attic heater", "unit_3": "kitchen speaker", "unit_4": "kitchen fan", "unit_5": "balcony speaker", "unit_6": "garage heater", "unit_7": "attic lamp", "unit_8": "bedroom speaker", "unit_9": "library lamp", "unit_10": "bedroom fan", "unit_11": "hall fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garage\",\"lamp\",true,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_8": 0, "unit_11": 0, "unit_4": 0, "unit_3": 0, "unit_7": 0, "unit_9": 0, "unit_1": 1, "unit_6": 0, "unit_10": 0, "unit_0": 0, "unit_5": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:86", "state_id": "smart_home_v2:train:86", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the fan in the garden. Authorization=no; people present=3; time=20:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library heater", "unit_1": "bathroom lamp", "unit_2": "garden fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garden\",\"fan\",false,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 1, "unit_1": 0, "unit_0": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:90", "state_id": "smart_home_v2:train:90", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the fan in the bedroom. Authorization=yes; people present=0; time=14:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "attic lamp", "unit_1": "bedroom fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bedroom\",\"fan\",true,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 0.97, "1": 0.03, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:88", "state_id": "smart_home_v2:train:88", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the fan in the bedroom. Authorization=yes; people present=3; time=4:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony fan", "unit_1": "bedroom fan", "unit_2": "garage lamp", "unit_3": "living room fan", "unit_4": "balcony heater", "unit_5": "attic speaker", "unit_6": "kitchen lamp", "unit_7": "office heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bedroom\",\"fan\",true,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_7": 0, "unit_0": 0, "unit_2": 0, "unit_3": 0, "unit_6": 0, "unit_1": 1, "unit_4": 0, "unit_5": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 0.99, "1": 0, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:87", "state_id": "smart_home_v2:train:87", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the lamp in the dining room. Authorization=yes; people present=2; time=13:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "kitchen fan", "unit_1": "dining room heater", "unit_2": "balcony lamp", "unit_3": "dining room lamp", "unit_4": "balcony fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_3", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 1, "unit_4": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"dining room\",\"lamp\",true,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_3": 1, "unit_2": 0, "unit_1": 0, "unit_4": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:89", "state_id": "smart_home_v2:train:89", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the lamp in the balcony. Authorization=no; people present=0; time=14:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room fan", "unit_1": "balcony lamp", "unit_2": "hall lamp", "unit_3": "library lamp", "unit_4": "hall fan", "unit_5": "dining room lamp", "unit_6": "bedroom fan", "unit_7": "office lamp", "unit_8": "attic fan", "unit_9": "bathroom speaker", "unit_10": "balcony heater", "unit_11": "garage fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"balcony\",\"lamp\",false,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_5": 0, "unit_10": 0, "unit_9": 0, "unit_7": 0, "unit_4": 0, "unit_8": 0, "unit_6": 0, "unit_3": 0, "unit_0": 0, "unit_1": 1, "unit_2": 0, "unit_11": 0}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:85", "state_id": "smart_home_v2:train:85", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the fan in the hall. Authorization=yes; people present=1; time=10:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "hall fan", "unit_1": "bedroom speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"hall\",\"fan\",true,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_0": 1}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:92", "state_id": "smart_home_v2:train:92", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the speaker in the bathroom. Authorization=yes; people present=0; time=18:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "dining room speaker", "unit_1": "bathroom speaker", "unit_2": "attic fan", "unit_3": "garage speaker", "unit_4": "kitchen heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0, "unit_3": 0, "unit_4": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bathroom\",\"speaker\",true,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_2": 0, "unit_4": 0, "unit_3": 0, "unit_1": 1}, "execute": {"false": 0.07999999999999996, "true": 0.92}, "risk": {"0": 0.71, "1": 0.27, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:91", "state_id": "smart_home_v2:train:91", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the fan in the living room. Authorization=no; people present=1; time=15:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "kitchen heater", "unit_1": "bathroom lamp", "unit_2": "living room fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"living room\",\"fan\",false,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:93", "state_id": "smart_home_v2:train:93", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the speaker in the hall. Authorization=yes; people present=1; time=0:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "hall speaker", "unit_1": "kitchen speaker", "unit_2": "bathroom speaker", "unit_3": "kitchen fan", "unit_4": "bedroom fan", "unit_5": "hall fan", "unit_6": "living room speaker", "unit_7": "bedroom lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"hall\",\"speaker\",true,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_7": 0, "unit_3": 0, "unit_2": 0, "unit_5": 0, "unit_6": 0, "unit_4": 0, "unit_1": 0, "unit_0": 1}, "execute": {"false": 0.06000000000000005, "true": 0.94}, "risk": {"0": 0.96, "1": 0.02, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:94", "state_id": "smart_home_v2:train:94", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the heater in the garden. Authorization=yes; people present=1; time=18:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom heater", "unit_1": "dining room speaker", "unit_2": "attic fan", "unit_3": "library heater", "unit_4": "bathroom heater", "unit_5": "library speaker", "unit_6": "balcony fan", "unit_7": "garden heater", "unit_8": "attic lamp", "unit_9": "living room speaker", "unit_10": "bathroom speaker", "unit_11": "bedroom fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_7", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 1, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garden\",\"heater\",true,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_4": 0, "unit_6": 0, "unit_9": 0, "unit_0": 0, "unit_7": 1, "unit_2": 0, "unit_10": 0, "unit_3": 0, "unit_8": 0, "unit_1": 0, "unit_11": 0, "unit_5": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:97", "state_id": "smart_home_v2:train:97", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the fan in the hall. Authorization=no; people present=2; time=3:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "kitchen fan", "unit_1": "hall fan", "unit_2": "attic lamp", "unit_3": "garage speaker", "unit_4": "library heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0, "unit_3": 0, "unit_4": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"hall\",\"fan\",false,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_3": 0, "unit_2": 0, "unit_1": 1, "unit_0": 0, "unit_4": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:96", "state_id": "smart_home_v2:train:96", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the heater in the office. Authorization=yes; people present=0; time=7:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "hall speaker", "unit_1": "office heater", "unit_2": "attic speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 1}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 1, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"office\",\"heater\",true,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 1, "unit_0": 0, "unit_2": 0}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0, "1": 1, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:99", "state_id": "smart_home_v2:train:99", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the lamp in the kitchen. Authorization=yes; people present=1; time=20:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room lamp", "unit_1": "hall lamp", "unit_2": "library fan", "unit_3": "office fan", "unit_4": "kitchen lamp", "unit_5": "hall speaker", "unit_6": "garden fan", "unit_7": "balcony lamp", "unit_8": "hall fan", "unit_9": "living room speaker", "unit_10": "library lamp", "unit_11": "garage lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_4", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 1, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"kitchen\",\"lamp\",true,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_7": 0, "unit_9": 0, "unit_3": 0, "unit_1": 0, "unit_6": 0, "unit_11": 0, "unit_8": 0, "unit_0": 0, "unit_10": 0, "unit_4": 1, "unit_5": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:98", "state_id": "smart_home_v2:train:98", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the speaker in the balcony. Authorization=no; people present=3; time=9:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom heater", "unit_1": "living room heater", "unit_2": "kitchen heater", "unit_3": "bedroom lamp", "unit_4": "bedroom fan", "unit_5": "attic lamp", "unit_6": "garage lamp", "unit_7": "balcony speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_7", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"balcony\",\"speaker\",false,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_4": 0, "unit_7": 1, "unit_6": 0, "unit_1": 0, "unit_3": 0, "unit_0": 0, "unit_5": 0, "unit_2": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:95", "state_id": "smart_home_v2:train:95", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the speaker in the bathroom. Authorization=no; people present=2; time=13:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "dining room fan", "unit_1": "bathroom speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bathroom\",\"speaker\",false,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:100", "state_id": "smart_home_v2:train:100", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the fan in the dining room. Authorization=no; people present=3; time=19:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "dining room fan", "unit_1": "bathroom speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"dining room\",\"fan\",false,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_0": 1}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.02, "1": 0, "2": 0.98}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:101", "state_id": "smart_home_v2:train:101", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the speaker in the attic. Authorization=no; people present=3; time=23:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "attic speaker", "unit_1": "hall speaker", "unit_2": "attic heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"attic\",\"speaker\",false,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_2": 0, "unit_1": 0}, "execute": {"false": 0.96, "true": 0.04}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:102", "state_id": "smart_home_v2:train:102", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the fan in the garage. Authorization=no; people present=1; time=16:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "hall fan", "unit_1": "dining room speaker", "unit_2": "kitchen fan", "unit_3": "garage fan", "unit_4": "balcony heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_3", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 1, "unit_4": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garage\",\"fan\",false,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_4": 0, "unit_3": 1, "unit_0": 0, "unit_1": 0}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:104", "state_id": "smart_home_v2:train:104", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the heater in the office. Authorization=yes; people present=2; time=19:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "kitchen speaker", "unit_1": "library lamp", "unit_2": "library heater", "unit_3": "dining room heater", "unit_4": "living room speaker", "unit_5": "office heater", "unit_6": "bathroom speaker", "unit_7": "library fan", "unit_8": "living room lamp", "unit_9": "office lamp", "unit_10": "bedroom heater", "unit_11": "hall speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_5", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 1, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"office\",\"heater\",true,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_4": 0, "unit_7": 0, "unit_9": 0, "unit_6": 0, "unit_1": 0, "unit_3": 0, "unit_11": 0, "unit_8": 0, "unit_10": 0, "unit_0": 0, "unit_2": 0, "unit_5": 1}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:103", "state_id": "smart_home_v2:train:103", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the speaker in the garage. Authorization=no; people present=2; time=17:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garden lamp", "unit_1": "hall lamp", "unit_2": "library heater", "unit_3": "office fan", "unit_4": "living room heater", "unit_5": "garage speaker", "unit_6": "bathroom speaker", "unit_7": "attic fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_5", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 1, "unit_6": 0, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garage\",\"speaker\",false,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_7": 0, "unit_2": 0, "unit_3": 0, "unit_5": 1, "unit_6": 0, "unit_4": 0, "unit_1": 0, "unit_0": 0}, "execute": {"false": 0.96, "true": 0.04}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:105", "state_id": "smart_home_v2:train:105", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the heater in the hall. Authorization=no; people present=1; time=0:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "hall heater", "unit_1": "balcony speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"hall\",\"heater\",false,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_0": 1}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:106", "state_id": "smart_home_v2:train:106", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the speaker in the bedroom. Authorization=no; people present=3; time=17:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom speaker", "unit_1": "attic heater", "unit_2": "garage speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bedroom\",\"speaker\",false,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_1": 0, "unit_0": 1}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:108", "state_id": "smart_home_v2:train:108", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the speaker in the hall. Authorization=yes; people present=0; time=22:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "dining room heater", "unit_1": "hall heater", "unit_2": "hall lamp", "unit_3": "kitchen speaker", "unit_4": "kitchen fan", "unit_5": "hall speaker", "unit_6": "dining room speaker", "unit_7": "attic fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_5", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 1, "unit_6": 0, "unit_7": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"hall\",\"speaker\",true,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_5": 1, "unit_0": 0, "unit_2": 0, "unit_4": 0, "unit_7": 0, "unit_3": 0, "unit_6": 0}, "execute": {"false": 0.07999999999999996, "true": 0.92}, "risk": {"0": 0.79, "1": 0.19, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:109", "state_id": "smart_home_v2:train:109", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the fan in the dining room. Authorization=no; people present=0; time=15:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "dining room fan", "unit_1": "bathroom fan", "unit_2": "living room speaker", "unit_3": "office speaker", "unit_4": "office lamp", "unit_5": "living room fan", "unit_6": "garage heater", "unit_7": "garage lamp", "unit_8": "bedroom lamp", "unit_9": "living room lamp", "unit_10": "kitchen speaker", "unit_11": "living room heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"dining room\",\"fan\",false,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_6": 0, "unit_8": 0, "unit_10": 0, "unit_3": 0, "unit_9": 0, "unit_1": 0, "unit_11": 0, "unit_0": 1, "unit_5": 0, "unit_2": 0, "unit_4": 0, "unit_7": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:110", "state_id": "smart_home_v2:train:110", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the heater in the library. Authorization=yes; people present=0; time=5:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library heater", "unit_1": "hall lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 1}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 1, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"library\",\"heater\",true,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_0": 1}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 1, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:111", "state_id": "smart_home_v2:train:111", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the speaker in the hall. Authorization=no; people present=2; time=21:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garden speaker", "unit_1": "bathroom speaker", "unit_2": "hall speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"hall\",\"speaker\",false,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 1, "unit_1": 0, "unit_0": 0}, "execute": {"false": 0.96, "true": 0.04}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:115", "state_id": "smart_home_v2:train:115", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the speaker in the balcony. Authorization=yes; people present=3; time=0:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony speaker", "unit_1": "library heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"balcony\",\"speaker\",true,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0.07999999999999996, "true": 0.92}, "risk": {"0": 0.97, "1": 0.01, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:113", "state_id": "smart_home_v2:train:113", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the fan in the garden. Authorization=no; people present=0; time=8:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bathroom lamp", "unit_1": "garden fan", "unit_2": "hall heater", "unit_3": "library fan", "unit_4": "attic heater", "unit_5": "bathroom speaker", "unit_6": "bathroom heater", "unit_7": "kitchen fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garden\",\"fan\",false,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_5": 0, "unit_0": 0, "unit_4": 0, "unit_3": 0, "unit_2": 0, "unit_1": 1, "unit_6": 0, "unit_7": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.02, "1": 0, "2": 0.98}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:112", "state_id": "smart_home_v2:train:112", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the speaker in the office. Authorization=yes; people present=0; time=14:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony heater", "unit_1": "garage speaker", "unit_2": "balcony speaker", "unit_3": "hall speaker", "unit_4": "office speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_4", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"office\",\"speaker\",true,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_1": 0, "unit_4": 1, "unit_0": 0, "unit_3": 0}, "execute": {"false": 0.07999999999999996, "true": 0.92}, "risk": {"0": 0.49, "1": 0.5, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:114", "state_id": "smart_home_v2:train:114", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the fan in the garden. Authorization=no; people present=2; time=18:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bathroom heater", "unit_1": "balcony lamp", "unit_2": "kitchen lamp", "unit_3": "garage heater", "unit_4": "library heater", "unit_5": "bathroom lamp", "unit_6": "dining room heater", "unit_7": "living room fan", "unit_8": "bathroom fan", "unit_9": "living room heater", "unit_10": "garden fan", "unit_11": "garage lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_10", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 1, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garden\",\"fan\",false,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_7": 0, "unit_9": 0, "unit_10": 1, "unit_2": 0, "unit_0": 0, "unit_4": 0, "unit_1": 0, "unit_5": 0, "unit_8": 0, "unit_3": 0, "unit_6": 0, "unit_11": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:116", "state_id": "smart_home_v2:train:116", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the heater in the garage. Authorization=yes; people present=0; time=14:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "office fan", "unit_1": "hall fan", "unit_2": "garage heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": false, "risk": 1}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 1, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garage\",\"heater\",true,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 1, "unit_1": 0, "unit_0": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 1, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:107", "state_id": "smart_home_v2:train:107", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the fan in the library. Authorization=no; people present=1; time=0:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "attic fan", "unit_1": "office fan", "unit_2": "hall lamp", "unit_3": "living room heater", "unit_4": "library fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_4", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"library\",\"fan\",false,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_2": 0, "unit_4": 1, "unit_1": 0, "unit_3": 0}, "execute": {"false": 0.96, "true": 0.04}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:117", "state_id": "smart_home_v2:train:117", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the heater in the bedroom. Authorization=no; people present=2; time=22:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library lamp", "unit_1": "bedroom heater", "unit_2": "garden lamp", "unit_3": "bathroom speaker", "unit_4": "garden fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0, "unit_3": 0, "unit_4": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bedroom\",\"heater\",false,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_3": 0, "unit_4": 0, "unit_2": 0, "unit_1": 1}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:118", "state_id": "smart_home_v2:train:118", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the fan in the dining room. Authorization=yes; people present=0; time=10:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bathroom speaker", "unit_1": "attic fan", "unit_2": "garden lamp", "unit_3": "hall speaker", "unit_4": "garage fan", "unit_5": "dining room fan", "unit_6": "kitchen lamp", "unit_7": "garage heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_5", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 1, "unit_6": 0, "unit_7": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"dining room\",\"fan\",true,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_3": 0, "unit_7": 0, "unit_6": 0, "unit_2": 0, "unit_0": 0, "unit_5": 1, "unit_4": 0, "unit_1": 0}, "execute": {"false": 0.050000000000000044, "true": 0.95}, "risk": {"0": 0.95, "1": 0.05, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:119", "state_id": "smart_home_v2:train:119", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the heater in the garden. Authorization=no; people present=3; time=6:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garage lamp", "unit_1": "library speaker", "unit_2": "living room lamp", "unit_3": "attic heater", "unit_4": "garage fan", "unit_5": "office heater", "unit_6": "garden lamp", "unit_7": "living room speaker", "unit_8": "garden speaker", "unit_9": "attic fan", "unit_10": "bathroom heater", "unit_11": "garden heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_11", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garden\",\"heater\",false,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_7": 0, "unit_9": 0, "unit_10": 0, "unit_4": 0, "unit_0": 0, "unit_2": 0, "unit_11": 1, "unit_5": 0, "unit_8": 0, "unit_6": 0, "unit_3": 0, "unit_1": 0}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:121", "state_id": "smart_home_v2:train:121", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the heater in the bathroom. Authorization=no; people present=3; time=20:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "dining room fan", "unit_1": "bathroom heater", "unit_2": "garage speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bathroom\",\"heater\",false,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:120", "state_id": "smart_home_v2:train:120", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the fan in the garage. Authorization=yes; people present=0; time=13:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony heater", "unit_1": "garage fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garage\",\"fan\",true,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 1, "unit_0": 0}, "execute": {"false": 0.050000000000000044, "true": 0.95}, "risk": {"0": 0.84, "1": 0.15, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:122", "state_id": "smart_home_v2:train:122", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the lamp in the bedroom. Authorization=no; people present=1; time=3:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room fan", "unit_1": "bathroom fan", "unit_2": "bedroom lamp", "unit_3": "bathroom speaker", "unit_4": "living room lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1, "unit_3": 0, "unit_4": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bedroom\",\"lamp\",false,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_4": 0, "unit_0": 0, "unit_1": 0, "unit_2": 1, "unit_3": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:124", "state_id": "smart_home_v2:train:124", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the lamp in the kitchen. Authorization=no; people present=1; time=8:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony fan", "unit_1": "attic lamp", "unit_2": "kitchen speaker", "unit_3": "hall fan", "unit_4": "hall heater", "unit_5": "office heater", "unit_6": "bathroom speaker", "unit_7": "library fan", "unit_8": "living room heater", "unit_9": "hall speaker", "unit_10": "bathroom fan", "unit_11": "kitchen lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_11", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"kitchen\",\"lamp\",false,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_5": 0, "unit_10": 0, "unit_0": 0, "unit_11": 1, "unit_3": 0, "unit_2": 0, "unit_6": 0, "unit_1": 0, "unit_9": 0, "unit_4": 0, "unit_7": 0, "unit_8": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:123", "state_id": "smart_home_v2:train:123", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the fan in the kitchen. Authorization=no; people present=1; time=4:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "kitchen speaker", "unit_1": "garden lamp", "unit_2": "library speaker", "unit_3": "hall fan", "unit_4": "office fan", "unit_5": "kitchen fan", "unit_6": "bedroom fan", "unit_7": "hall lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_5", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 1, "unit_6": 0, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"kitchen\",\"fan\",false,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_1": 0, "unit_6": 0, "unit_7": 0, "unit_0": 0, "unit_3": 0, "unit_5": 1, "unit_4": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:125", "state_id": "smart_home_v2:train:125", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the lamp in the attic. Authorization=no; people present=3; time=21:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "attic lamp", "unit_1": "office heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"attic\",\"lamp\",false,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:128", "state_id": "smart_home_v2:train:128", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the heater in the dining room. Authorization=yes; people present=1; time=4:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bathroom heater", "unit_1": "bedroom heater", "unit_2": "dining room heater", "unit_3": "dining room speaker", "unit_4": "garage lamp", "unit_5": "living room heater", "unit_6": "living room lamp", "unit_7": "hall fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"dining room\",\"heater\",true,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_6": 0, "unit_5": 0, "unit_2": 1, "unit_7": 0, "unit_4": 0, "unit_3": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:127", "state_id": "smart_home_v2:train:127", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the heater in the dining room. Authorization=yes; people present=2; time=10:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "dining room heater", "unit_1": "kitchen fan", "unit_2": "hall heater", "unit_3": "living room lamp", "unit_4": "garden lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"dining room\",\"heater\",true,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_2": 0, "unit_3": 0, "unit_1": 0, "unit_4": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:130", "state_id": "smart_home_v2:train:130", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the lamp in the garage. Authorization=yes; people present=1; time=17:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "dining room heater", "unit_1": "garage lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garage\",\"lamp\",true,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 1, "unit_0": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:129", "state_id": "smart_home_v2:train:129", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the fan in the bedroom. Authorization=yes; people present=2; time=22:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "attic heater", "unit_1": "dining room speaker", "unit_2": "balcony speaker", "unit_3": "dining room fan", "unit_4": "balcony lamp", "unit_5": "living room lamp", "unit_6": "garage heater", "unit_7": "kitchen speaker", "unit_8": "garden lamp", "unit_9": "garage speaker", "unit_10": "bedroom fan", "unit_11": "bedroom lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_10", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 1, "unit_11": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bedroom\",\"fan\",true,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_0": 0, "unit_11": 0, "unit_4": 0, "unit_3": 0, "unit_7": 0, "unit_9": 0, "unit_1": 0, "unit_6": 0, "unit_10": 1, "unit_8": 0, "unit_5": 0}, "execute": {"false": 0.040000000000000036, "true": 0.96}, "risk": {"0": 0.98, "1": 0, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:126", "state_id": "smart_home_v2:train:126", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the heater in the balcony. Authorization=yes; people present=2; time=20:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony heater", "unit_1": "office heater", "unit_2": "bedroom speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"balcony\",\"heater\",true,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:132", "state_id": "smart_home_v2:train:132", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the heater in the hall. Authorization=yes; people present=3; time=10:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room speaker", "unit_1": "kitchen lamp", "unit_2": "dining room heater", "unit_3": "hall heater", "unit_4": "library speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_3", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 1, "unit_4": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"hall\",\"heater\",true,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_0": 0, "unit_1": 0, "unit_4": 0, "unit_3": 1}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:133", "state_id": "smart_home_v2:train:133", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the speaker in the attic. Authorization=no; people present=2; time=11:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "attic fan", "unit_1": "bedroom speaker", "unit_2": "bathroom fan", "unit_3": "attic lamp", "unit_4": "living room lamp", "unit_5": "attic speaker", "unit_6": "dining room fan", "unit_7": "bathroom heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_5", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 1, "unit_6": 0, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"attic\",\"speaker\",false,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_7": 0, "unit_6": 0, "unit_2": 0, "unit_3": 0, "unit_5": 1, "unit_4": 0, "unit_1": 0}, "execute": {"false": 0.96, "true": 0.04}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:131", "state_id": "smart_home_v2:train:131", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the speaker in the library. Authorization=no; people present=2; time=3:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garage speaker", "unit_1": "bedroom lamp", "unit_2": "library speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"library\",\"speaker\",false,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_2": 1, "unit_1": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:134", "state_id": "smart_home_v2:train:134", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the lamp in the attic. Authorization=no; people present=2; time=23:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garden heater", "unit_1": "living room fan", "unit_2": "attic lamp", "unit_3": "office speaker", "unit_4": "hall heater", "unit_5": "garage speaker", "unit_6": "living room heater", "unit_7": "garage fan", "unit_8": "living room lamp", "unit_9": "balcony lamp", "unit_10": "bathroom speaker", "unit_11": "bedroom heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"attic\",\"lamp\",false,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 1, "unit_3": 0, "unit_8": 0, "unit_9": 0, "unit_11": 0, "unit_0": 0, "unit_5": 0, "unit_10": 0, "unit_4": 0, "unit_6": 0, "unit_1": 0, "unit_7": 0}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:136", "state_id": "smart_home_v2:train:136", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the lamp in the kitchen. Authorization=no; people present=3; time=11:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "kitchen lamp", "unit_1": "balcony fan", "unit_2": "library fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"kitchen\",\"lamp\",false,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_2": 0, "unit_1": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:137", "state_id": "smart_home_v2:train:137", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the lamp in the living room. Authorization=yes; people present=3; time=12:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room lamp", "unit_1": "bathroom heater", "unit_2": "hall fan", "unit_3": "garage heater", "unit_4": "garage fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"living room\",\"lamp\",true,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_3": 0, "unit_1": 0, "unit_0": 1, "unit_4": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:135", "state_id": "smart_home_v2:train:135", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the heater in the bathroom. Authorization=no; people present=0; time=19:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bathroom heater", "unit_1": "library heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bathroom\",\"heater\",false,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:138", "state_id": "smart_home_v2:train:138", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the lamp in the living room. Authorization=yes; people present=2; time=11:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "attic heater", "unit_1": "bathroom speaker", "unit_2": "bedroom fan", "unit_3": "balcony speaker", "unit_4": "living room heater", "unit_5": "living room lamp", "unit_6": "garage fan", "unit_7": "hall lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_5", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 1, "unit_6": 0, "unit_7": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"living room\",\"lamp\",true,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_4": 0, "unit_5": 1, "unit_2": 0, "unit_7": 0, "unit_1": 0, "unit_3": 0, "unit_6": 0, "unit_0": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:139", "state_id": "smart_home_v2:train:139", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the fan in the office. Authorization=no; people present=1; time=0:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library speaker", "unit_1": "kitchen lamp", "unit_2": "kitchen heater", "unit_3": "garden fan", "unit_4": "hall lamp", "unit_5": "attic fan", "unit_6": "library lamp", "unit_7": "hall heater", "unit_8": "office fan", "unit_9": "garden speaker", "unit_10": "bedroom lamp", "unit_11": "garage heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_8", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 1, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"office\",\"fan\",false,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_3": 0, "unit_8": 1, "unit_4": 0, "unit_2": 0, "unit_9": 0, "unit_11": 0, "unit_0": 0, "unit_6": 0, "unit_10": 0, "unit_7": 0, "unit_5": 0, "unit_1": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:140", "state_id": "smart_home_v2:train:140", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the fan in the balcony. Authorization=no; people present=1; time=21:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garage fan", "unit_1": "balcony fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"balcony\",\"fan\",false,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 1, "unit_0": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:141", "state_id": "smart_home_v2:train:141", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the fan in the library. Authorization=no; people present=2; time=12:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "office heater", "unit_1": "library fan", "unit_2": "bathroom lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"library\",\"fan\",false,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_2": 0, "unit_1": 1}, "execute": {"false": 0.96, "true": 0.04}, "risk": {"0": 0.02, "1": 0, "2": 0.98}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:142", "state_id": "smart_home_v2:train:142", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the heater in the library. Authorization=yes; people present=3; time=14:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room fan", "unit_1": "bathroom heater", "unit_2": "hall speaker", "unit_3": "library heater", "unit_4": "hall lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_3", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 1, "unit_4": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"library\",\"heater\",true,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_2": 0, "unit_3": 1, "unit_1": 0, "unit_4": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:144", "state_id": "smart_home_v2:train:144", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the fan in the garden. Authorization=yes; people present=1; time=16:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room lamp", "unit_1": "kitchen heater", "unit_2": "garage heater", "unit_3": "bedroom heater", "unit_4": "bathroom speaker", "unit_5": "living room speaker", "unit_6": "balcony speaker", "unit_7": "kitchen fan", "unit_8": "library lamp", "unit_9": "kitchen speaker", "unit_10": "library fan", "unit_11": "garden fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_11", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garden\",\"fan\",true,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_5": 0, "unit_9": 0, "unit_10": 0, "unit_7": 0, "unit_4": 0, "unit_8": 0, "unit_6": 0, "unit_3": 0, "unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_11": 1}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:143", "state_id": "smart_home_v2:train:143", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the fan in the dining room. Authorization=yes; people present=1; time=0:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library fan", "unit_1": "bathroom speaker", "unit_2": "balcony fan", "unit_3": "kitchen fan", "unit_4": "hall heater", "unit_5": "bedroom lamp", "unit_6": "dining room fan", "unit_7": "balcony heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_6", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 1, "unit_7": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"dining room\",\"fan\",true,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_4": 0, "unit_7": 0, "unit_3": 0, "unit_6": 1, "unit_1": 0, "unit_5": 0, "unit_2": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:145", "state_id": "smart_home_v2:train:145", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the heater in the hall. Authorization=no; people present=3; time=6:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library heater", "unit_1": "hall heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"hall\",\"heater\",false,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:147", "state_id": "smart_home_v2:train:147", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the fan in the balcony. Authorization=no; people present=0; time=3:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "kitchen lamp", "unit_1": "garage heater", "unit_2": "living room heater", "unit_3": "balcony fan", "unit_4": "kitchen heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_3", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 1, "unit_4": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"balcony\",\"fan\",false,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_4": 0, "unit_1": 0, "unit_3": 1, "unit_0": 0, "unit_2": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:149", "state_id": "smart_home_v2:train:149", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the fan in the dining room. Authorization=no; people present=2; time=4:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room heater", "unit_1": "bedroom fan", "unit_2": "hall lamp", "unit_3": "library heater", "unit_4": "dining room fan", "unit_5": "library lamp", "unit_6": "bathroom heater", "unit_7": "office lamp", "unit_8": "kitchen fan", "unit_9": "hall heater", "unit_10": "garage lamp", "unit_11": "living room speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_4", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 1, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"dining room\",\"fan\",false,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_5": 0, "unit_0": 0, "unit_4": 1, "unit_8": 0, "unit_1": 0, "unit_10": 0, "unit_2": 0, "unit_11": 0, "unit_9": 0, "unit_6": 0, "unit_7": 0, "unit_3": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:146", "state_id": "smart_home_v2:train:146", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the lamp in the bedroom. Authorization=yes; people present=3; time=17:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom fan", "unit_1": "library speaker", "unit_2": "bedroom lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bedroom\",\"lamp\",true,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 1, "unit_1": 0, "unit_0": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:150", "state_id": "smart_home_v2:train:150", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the lamp in the garden. Authorization=no; people present=0; time=9:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "office speaker", "unit_1": "garden lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garden\",\"lamp\",false,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 1, "unit_0": 0}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:153", "state_id": "smart_home_v2:train:153", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the lamp in the hall. Authorization=yes; people present=2; time=2:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom fan", "unit_1": "kitchen lamp", "unit_2": "dining room lamp", "unit_3": "hall lamp", "unit_4": "bathroom lamp", "unit_5": "garden fan", "unit_6": "balcony fan", "unit_7": "living room heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_3", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 1, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"hall\",\"lamp\",true,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_7": 0, "unit_3": 1, "unit_1": 0, "unit_6": 0, "unit_0": 0, "unit_5": 0, "unit_4": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:151", "state_id": "smart_home_v2:train:151", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the fan in the library. Authorization=yes; people present=2; time=14:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony heater", "unit_1": "dining room fan", "unit_2": "library fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"library\",\"fan\",true,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 1, "unit_1": 0, "unit_0": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:154", "state_id": "smart_home_v2:train:154", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the speaker in the bathroom. Authorization=no; people present=0; time=16:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bathroom speaker", "unit_1": "library speaker", "unit_2": "garden speaker", "unit_3": "balcony heater", "unit_4": "living room fan", "unit_5": "office lamp", "unit_6": "dining room heater", "unit_7": "balcony lamp", "unit_8": "bathroom lamp", "unit_9": "living room lamp", "unit_10": "garage speaker", "unit_11": "attic speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bathroom\",\"speaker\",false,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_7": 0, "unit_0": 1, "unit_5": 0, "unit_10": 0, "unit_8": 0, "unit_3": 0, "unit_11": 0, "unit_4": 0, "unit_2": 0, "unit_6": 0, "unit_9": 0}, "execute": {"false": 0.96, "true": 0.04}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:152", "state_id": "smart_home_v2:train:152", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the fan in the garage. Authorization=yes; people present=1; time=13:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garage fan", "unit_1": "garden speaker", "unit_2": "dining room lamp", "unit_3": "bedroom speaker", "unit_4": "attic heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garage\",\"fan\",true,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_4": 0, "unit_0": 1, "unit_1": 0, "unit_2": 0, "unit_3": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:156", "state_id": "smart_home_v2:train:156", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the fan in the living room. Authorization=no; people present=0; time=19:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room fan", "unit_1": "garage speaker", "unit_2": "garage fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"living room\",\"fan\",false,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_0": 1, "unit_2": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:148", "state_id": "smart_home_v2:train:148", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the heater in the garage. Authorization=no; people present=1; time=13:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garage speaker", "unit_1": "dining room lamp", "unit_2": "garage heater", "unit_3": "dining room fan", "unit_4": "kitchen heater", "unit_5": "library heater", "unit_6": "dining room heater", "unit_7": "garden fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garage\",\"heater\",false,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_3": 0, "unit_4": 0, "unit_2": 1, "unit_0": 0, "unit_6": 0, "unit_7": 0, "unit_5": 0, "unit_1": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:157", "state_id": "smart_home_v2:train:157", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the lamp in the garden. Authorization=yes; people present=0; time=5:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "dining room heater", "unit_1": "garden lamp", "unit_2": "garage heater", "unit_3": "attic speaker", "unit_4": "balcony lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0, "unit_3": 0, "unit_4": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garden\",\"lamp\",true,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_3": 0, "unit_4": 0, "unit_2": 0}, "execute": {"false": 0.040000000000000036, "true": 0.96}, "risk": {"0": 0.96, "1": 0.02, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:155", "state_id": "smart_home_v2:train:155", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the heater in the library. Authorization=yes; people present=2; time=16:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "hall lamp", "unit_1": "library heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"library\",\"heater\",true,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 1, "unit_0": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:160", "state_id": "smart_home_v2:train:160", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the heater in the balcony. Authorization=yes; people present=3; time=6:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony heater", "unit_1": "balcony lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"balcony\",\"heater\",true,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:161", "state_id": "smart_home_v2:train:161", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the lamp in the dining room. Authorization=no; people present=0; time=18:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "dining room lamp", "unit_1": "office heater", "unit_2": "attic speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"dining room\",\"lamp\",false,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_0": 1, "unit_2": 0}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:158", "state_id": "smart_home_v2:train:158", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the heater in the kitchen. Authorization=no; people present=3; time=6:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "hall speaker", "unit_1": "garage heater", "unit_2": "living room fan", "unit_3": "balcony fan", "unit_4": "office speaker", "unit_5": "kitchen heater", "unit_6": "attic speaker", "unit_7": "garage speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_5", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 1, "unit_6": 0, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"kitchen\",\"heater\",false,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_4": 0, "unit_3": 0, "unit_7": 0, "unit_1": 0, "unit_6": 0, "unit_0": 0, "unit_5": 1}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:162", "state_id": "smart_home_v2:train:162", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the lamp in the library. Authorization=no; people present=1; time=3:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "dining room heater", "unit_1": "dining room speaker", "unit_2": "living room speaker", "unit_3": "bedroom fan", "unit_4": "library lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_4", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"library\",\"lamp\",false,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_0": 0, "unit_4": 1, "unit_1": 0, "unit_3": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:159", "state_id": "smart_home_v2:train:159", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the fan in the living room. Authorization=yes; people present=0; time=19:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "hall heater", "unit_1": "dining room heater", "unit_2": "kitchen speaker", "unit_3": "garden heater", "unit_4": "garage lamp", "unit_5": "dining room speaker", "unit_6": "kitchen fan", "unit_7": "garden speaker", "unit_8": "garden lamp", "unit_9": "dining room fan", "unit_10": "balcony heater", "unit_11": "living room fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_11", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"living room\",\"fan\",true,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_2": 0, "unit_4": 0, "unit_3": 0, "unit_5": 0, "unit_7": 0, "unit_1": 0, "unit_10": 0, "unit_8": 0, "unit_6": 0, "unit_11": 1, "unit_9": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:165", "state_id": "smart_home_v2:train:165", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the speaker in the library. Authorization=yes; people present=0; time=1:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library speaker", "unit_1": "kitchen speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"library\",\"speaker\",true,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_0": 1}, "execute": {"false": 0.08999999999999997, "true": 0.91}, "risk": {"0": 0.68, "1": 0.31, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:163", "state_id": "smart_home_v2:train:163", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the fan in the hall. Authorization=no; people present=0; time=7:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garage speaker", "unit_1": "hall fan", "unit_2": "library lamp", "unit_3": "garage fan", "unit_4": "attic heater", "unit_5": "balcony speaker", "unit_6": "office heater", "unit_7": "bathroom lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"hall\",\"fan\",false,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_3": 0, "unit_6": 0, "unit_2": 0, "unit_0": 0, "unit_4": 0, "unit_5": 0, "unit_1": 1, "unit_7": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:166", "state_id": "smart_home_v2:train:166", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the speaker in the living room. Authorization=no; people present=1; time=11:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room speaker", "unit_1": "bathroom speaker", "unit_2": "garage lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"living room\",\"speaker\",false,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_1": 0, "unit_0": 1}, "execute": {"false": 0.96, "true": 0.04}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:167", "state_id": "smart_home_v2:train:167", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the speaker in the hall. Authorization=yes; people present=3; time=20:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "office heater", "unit_1": "office speaker", "unit_2": "hall speaker", "unit_3": "attic heater", "unit_4": "balcony speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1, "unit_3": 0, "unit_4": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"hall\",\"speaker\",true,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_3": 0, "unit_2": 1, "unit_1": 0, "unit_4": 0, "unit_0": 0}, "execute": {"false": 0.09999999999999998, "true": 0.9}, "risk": {"0": 0.97, "1": 0.01, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:168", "state_id": "smart_home_v2:train:168", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the speaker in the bathroom. Authorization=yes; people present=1; time=10:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garden speaker", "unit_1": "hall speaker", "unit_2": "dining room speaker", "unit_3": "hall lamp", "unit_4": "living room lamp", "unit_5": "hall fan", "unit_6": "attic lamp", "unit_7": "bathroom speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_7", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bathroom\",\"speaker\",true,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_0": 0, "unit_5": 0, "unit_3": 0, "unit_4": 0, "unit_2": 0, "unit_6": 0, "unit_7": 1}, "execute": {"false": 0.06000000000000005, "true": 0.94}, "risk": {"0": 0.99, "1": 0, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:169", "state_id": "smart_home_v2:train:169", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the lamp in the balcony. Authorization=no; people present=1; time=5:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room heater", "unit_1": "garden speaker", "unit_2": "bedroom lamp", "unit_3": "living room speaker", "unit_4": "living room fan", "unit_5": "garden heater", "unit_6": "attic speaker", "unit_7": "dining room fan", "unit_8": "balcony lamp", "unit_9": "office speaker", "unit_10": "bedroom speaker", "unit_11": "hall heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_8", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 1, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"balcony\",\"lamp\",false,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_7": 0, "unit_0": 0, "unit_5": 0, "unit_10": 0, "unit_8": 1, "unit_3": 0, "unit_11": 0, "unit_4": 0, "unit_2": 0, "unit_6": 0, "unit_9": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:171", "state_id": "smart_home_v2:train:171", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the heater in the balcony. Authorization=no; people present=1; time=21:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony lamp", "unit_1": "balcony heater", "unit_2": "library fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"balcony\",\"heater\",false,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_2": 0, "unit_1": 1}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:164", "state_id": "smart_home_v2:train:164", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the lamp in the hall. Authorization=yes; people present=1; time=22:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library heater", "unit_1": "garden lamp", "unit_2": "office heater", "unit_3": "hall heater", "unit_4": "bedroom heater", "unit_5": "balcony lamp", "unit_6": "hall lamp", "unit_7": "living room fan", "unit_8": "bathroom lamp", "unit_9": "garage heater", "unit_10": "bathroom fan", "unit_11": "bathroom heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_6", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 1, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"hall\",\"lamp\",true,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_5": 0, "unit_6": 1, "unit_0": 0, "unit_7": 0, "unit_2": 0, "unit_4": 0, "unit_11": 0, "unit_10": 0, "unit_9": 0, "unit_3": 0, "unit_8": 0, "unit_1": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:172", "state_id": "smart_home_v2:train:172", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the lamp in the garage. Authorization=no; people present=3; time=5:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garage lamp", "unit_1": "bathroom lamp", "unit_2": "dining room lamp", "unit_3": "living room speaker", "unit_4": "library fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garage\",\"lamp\",false,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_3": 0, "unit_0": 1, "unit_1": 0, "unit_2": 0, "unit_4": 0}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:173", "state_id": "smart_home_v2:train:173", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the speaker in the garage. Authorization=yes; people present=2; time=5:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room lamp", "unit_1": "dining room lamp", "unit_2": "garage speaker", "unit_3": "office fan", "unit_4": "hall speaker", "unit_5": "library fan", "unit_6": "balcony speaker", "unit_7": "living room speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garage\",\"speaker\",true,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 1, "unit_4": 0, "unit_3": 0, "unit_7": 0, "unit_1": 0, "unit_6": 0, "unit_0": 0, "unit_5": 0}, "execute": {"false": 0.050000000000000044, "true": 0.95}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:174", "state_id": "smart_home_v2:train:174", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the heater in the office. Authorization=no; people present=1; time=17:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room speaker", "unit_1": "dining room heater", "unit_2": "garden heater", "unit_3": "hall heater", "unit_4": "attic heater", "unit_5": "kitchen fan", "unit_6": "balcony heater", "unit_7": "office heater", "unit_8": "garage speaker", "unit_9": "library speaker", "unit_10": "bedroom heater", "unit_11": "dining room speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_7", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 1, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"office\",\"heater\",false,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_8": 0, "unit_3": 0, "unit_6": 0, "unit_4": 0, "unit_11": 0, "unit_2": 0, "unit_5": 0, "unit_9": 0, "unit_7": 1, "unit_0": 0, "unit_1": 0, "unit_10": 0}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:175", "state_id": "smart_home_v2:train:175", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the speaker in the garage. Authorization=yes; people present=0; time=17:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garage speaker", "unit_1": "living room fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garage\",\"speaker\",true,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0.050000000000000044, "true": 0.95}, "risk": {"0": 0.87, "1": 0.13, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:176", "state_id": "smart_home_v2:train:176", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the speaker in the kitchen. Authorization=no; people present=3; time=14:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "attic fan", "unit_1": "dining room speaker", "unit_2": "kitchen speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"kitchen\",\"speaker\",false,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 1, "unit_1": 0, "unit_0": 0}, "execute": {"false": 0.96, "true": 0.04}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:177", "state_id": "smart_home_v2:train:177", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the lamp in the living room. Authorization=no; people present=2; time=17:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room lamp", "unit_1": "attic fan", "unit_2": "garden heater", "unit_3": "balcony speaker", "unit_4": "bathroom fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"living room\",\"lamp\",false,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_3": 0, "unit_2": 0, "unit_4": 0, "unit_1": 0, "unit_0": 1}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:178", "state_id": "smart_home_v2:train:178", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the lamp in the library. Authorization=yes; people present=1; time=22:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garage lamp", "unit_1": "library fan", "unit_2": "library lamp", "unit_3": "bathroom lamp", "unit_4": "office heater", "unit_5": "living room heater", "unit_6": "kitchen heater", "unit_7": "hall speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"library\",\"lamp\",true,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 1, "unit_3": 0, "unit_7": 0, "unit_0": 0, "unit_5": 0, "unit_4": 0, "unit_1": 0, "unit_6": 0}, "execute": {"false": 0.040000000000000036, "true": 0.96}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:179", "state_id": "smart_home_v2:train:179", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the lamp in the office. Authorization=yes; people present=2; time=18:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony speaker", "unit_1": "office fan", "unit_2": "garage heater", "unit_3": "office lamp", "unit_4": "bathroom speaker", "unit_5": "bedroom heater", "unit_6": "garden heater", "unit_7": "bedroom lamp", "unit_8": "balcony fan", "unit_9": "living room heater", "unit_10": "garden fan", "unit_11": "dining room speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_3", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 1, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"office\",\"lamp\",true,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_10": 0, "unit_2": 0, "unit_1": 0, "unit_7": 0, "unit_8": 0, "unit_0": 0, "unit_5": 0, "unit_3": 1, "unit_11": 0, "unit_4": 0, "unit_9": 0, "unit_6": 0}, "execute": {"false": 0.040000000000000036, "true": 0.96}, "risk": {"0": 0.99, "1": 0, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:170", "state_id": "smart_home_v2:train:170", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the heater in the attic. Authorization=yes; people present=1; time=2:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garden heater", "unit_1": "attic heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"attic\",\"heater\",true,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 1, "unit_0": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:180", "state_id": "smart_home_v2:train:180", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the lamp in the library. Authorization=yes; people present=2; time=11:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library lamp", "unit_1": "hall heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"library\",\"lamp\",true,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_0": 1}, "execute": {"false": 0.040000000000000036, "true": 0.96}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:181", "state_id": "smart_home_v2:train:181", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the fan in the balcony. Authorization=yes; people present=1; time=16:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony fan", "unit_1": "dining room lamp", "unit_2": "hall heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"balcony\",\"fan\",true,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_1": 0, "unit_0": 1}, "execute": {"false": 0.050000000000000044, "true": 0.95}, "risk": {"0": 0.99, "1": 0, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:182", "state_id": "smart_home_v2:train:182", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the fan in the attic. Authorization=no; people present=3; time=3:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garden lamp", "unit_1": "bedroom lamp", "unit_2": "attic fan", "unit_3": "hall lamp", "unit_4": "attic speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1, "unit_3": 0, "unit_4": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"attic\",\"fan\",false,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_4": 0, "unit_2": 1, "unit_3": 0, "unit_1": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:184", "state_id": "smart_home_v2:train:184", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the fan in the garage. Authorization=no; people present=0; time=2:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bathroom heater", "unit_1": "hall heater", "unit_2": "office fan", "unit_3": "kitchen lamp", "unit_4": "living room speaker", "unit_5": "library heater", "unit_6": "garden fan", "unit_7": "kitchen fan", "unit_8": "garage lamp", "unit_9": "garage fan", "unit_10": "dining room heater", "unit_11": "bedroom lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_9", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 1, "unit_10": 0, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garage\",\"fan\",false,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_4": 0, "unit_7": 0, "unit_9": 1, "unit_3": 0, "unit_1": 0, "unit_6": 0, "unit_11": 0, "unit_8": 0, "unit_10": 0, "unit_0": 0, "unit_5": 0, "unit_2": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.02, "1": 0, "2": 0.98}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:187", "state_id": "smart_home_v2:train:187", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the speaker in the living room. Authorization=yes; people present=0; time=10:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room fan", "unit_1": "library lamp", "unit_2": "office lamp", "unit_3": "kitchen fan", "unit_4": "living room speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_4", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"living room\",\"speaker\",true,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_4": 1, "unit_1": 0, "unit_3": 0, "unit_0": 0, "unit_2": 0}, "execute": {"false": 0.06000000000000005, "true": 0.94}, "risk": {"0": 0.76, "1": 0.24, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:186", "state_id": "smart_home_v2:train:186", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the lamp in the living room. Authorization=yes; people present=0; time=16:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room lamp", "unit_1": "garage lamp", "unit_2": "hall fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"living room\",\"lamp\",true,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_1": 0, "unit_0": 1}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:183", "state_id": "smart_home_v2:train:183", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the heater in the attic. Authorization=no; people present=2; time=6:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "kitchen fan", "unit_1": "dining room lamp", "unit_2": "office heater", "unit_3": "hall heater", "unit_4": "bathroom heater", "unit_5": "attic heater", "unit_6": "garage speaker", "unit_7": "attic lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_5", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 1, "unit_6": 0, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"attic\",\"heater\",false,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_5": 1, "unit_7": 0, "unit_6": 0, "unit_1": 0, "unit_3": 0, "unit_0": 0, "unit_4": 0, "unit_2": 0}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:185", "state_id": "smart_home_v2:train:185", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the lamp in the bedroom. Authorization=yes; people present=2; time=18:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom lamp", "unit_1": "library speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bedroom\",\"lamp\",true,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_0": 1}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:188", "state_id": "smart_home_v2:train:188", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the heater in the garage. Authorization=no; people present=0; time=2:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garden fan", "unit_1": "bedroom fan", "unit_2": "hall fan", "unit_3": "garage heater", "unit_4": "bathroom speaker", "unit_5": "library heater", "unit_6": "library fan", "unit_7": "dining room heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_3", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 1, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garage\",\"heater\",false,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_4": 0, "unit_2": 0, "unit_5": 0, "unit_0": 0, "unit_6": 0, "unit_7": 0, "unit_3": 1}, "execute": {"false": 0.99, "true": 0.01}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:189", "state_id": "smart_home_v2:train:189", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the speaker in the library. Authorization=yes; people present=2; time=11:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony heater", "unit_1": "office fan", "unit_2": "garage heater", "unit_3": "hall speaker", "unit_4": "dining room fan", "unit_5": "hall fan", "unit_6": "library heater", "unit_7": "library lamp", "unit_8": "dining room lamp", "unit_9": "attic lamp", "unit_10": "library speaker", "unit_11": "balcony lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_10", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 1, "unit_11": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"library\",\"speaker\",true,2]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_7": 0, "unit_9": 0, "unit_10": 1, "unit_2": 0, "unit_0": 0, "unit_4": 0, "unit_1": 0, "unit_5": 0, "unit_8": 0, "unit_3": 0, "unit_6": 0, "unit_11": 0}, "execute": {"false": 0.09999999999999998, "true": 0.9}, "risk": {"0": 0.98, "1": 0, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:190", "state_id": "smart_home_v2:train:190", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the lamp in the office. Authorization=yes; people present=3; time=14:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "office lamp", "unit_1": "garden fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"office\",\"lamp\",true,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:193", "state_id": "smart_home_v2:train:193", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the fan in the garden. Authorization=yes; people present=3; time=14:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "office fan", "unit_1": "garden fan", "unit_2": "dining room fan", "unit_3": "kitchen heater", "unit_4": "attic fan", "unit_5": "garden lamp", "unit_6": "kitchen lamp", "unit_7": "library speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garden\",\"fan\",true,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_4": 0, "unit_7": 0, "unit_3": 0, "unit_6": 0, "unit_1": 1, "unit_5": 0, "unit_2": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:194", "state_id": "smart_home_v2:train:194", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the speaker in the bedroom. Authorization=yes; people present=3; time=19:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library heater", "unit_1": "kitchen fan", "unit_2": "kitchen lamp", "unit_3": "bedroom speaker", "unit_4": "library lamp", "unit_5": "attic speaker", "unit_6": "bathroom heater", "unit_7": "bedroom heater", "unit_8": "garage fan", "unit_9": "garden lamp", "unit_10": "balcony lamp", "unit_11": "living room lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_3", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 1, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bedroom\",\"speaker\",true,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_5": 0, "unit_8": 0, "unit_0": 0, "unit_2": 0, "unit_11": 0, "unit_9": 0, "unit_4": 0, "unit_7": 0, "unit_3": 1, "unit_10": 0, "unit_6": 0}, "execute": {"false": 0.06000000000000005, "true": 0.94}, "risk": {"0": 0.99, "1": 0, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:196", "state_id": "smart_home_v2:train:196", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the heater in the balcony. Authorization=no; people present=3; time=0:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library speaker", "unit_1": "bedroom fan", "unit_2": "balcony heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"balcony\",\"heater\",false,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_2": 1, "unit_0": 0}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:191", "state_id": "smart_home_v2:train:191", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the lamp in the garden. Authorization=yes; people present=1; time=4:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "attic fan", "unit_1": "garden lamp", "unit_2": "dining room fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garden\",\"lamp\",true,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:195", "state_id": "smart_home_v2:train:195", "family_id": "smart_home_v2", "split": "train", "state": "Request: off the speaker in the office. Authorization=no; people present=1; time=9:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "office speaker", "unit_1": "garden lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"office\",\"speaker\",false,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_0": 1}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.03, "1": 0, "2": 0.97}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:198", "state_id": "smart_home_v2:train:198", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the fan in the dining room. Authorization=yes; people present=3; time=11:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "attic heater", "unit_1": "dining room fan", "unit_2": "hall lamp", "unit_3": "garden speaker", "unit_4": "hall speaker", "unit_5": "library fan", "unit_6": "bathroom fan", "unit_7": "living room lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"dining room\",\"fan\",true,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_7": 0, "unit_2": 0, "unit_3": 0, "unit_5": 0, "unit_6": 0, "unit_4": 0, "unit_1": 1, "unit_0": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:0", "state_id": "catalog_lookup_v2:train:0", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00215; color=pink; size=large; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "pink, large", "candidate_1": "red, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00215\",\"pink\",\"large\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_0": 1}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:1", "state_id": "catalog_lookup_v2:train:1", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00254; color=gray; size=medium; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "orange, small", "candidate_1": "gray, medium", "candidate_2": "white, small", "candidate_3": "gray, large", "candidate_4": "pink, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00254\",\"gray\",\"medium\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 0, "candidate_2": 0, "candidate_4": 0, "candidate_1": 1, "candidate_3": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:192", "state_id": "smart_home_v2:train:192", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the lamp in the attic. Authorization=no; people present=1; time=16:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "attic lamp", "unit_1": "hall heater", "unit_2": "bathroom lamp", "unit_3": "kitchen lamp", "unit_4": "bedroom speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"attic\",\"lamp\",false,1]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_3": 0, "unit_2": 0, "unit_4": 0, "unit_1": 0, "unit_0": 1}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:199", "state_id": "smart_home_v2:train:199", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the lamp in the kitchen. Authorization=no; people present=0; time=20:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library lamp", "unit_1": "kitchen lamp", "unit_2": "bedroom speaker", "unit_3": "office lamp", "unit_4": "kitchen fan", "unit_5": "balcony fan", "unit_6": "office speaker", "unit_7": "hall speaker", "unit_8": "library heater", "unit_9": "bathroom heater", "unit_10": "bathroom speaker", "unit_11": "living room speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"kitchen\",\"lamp\",false,0]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_1": 1, "unit_6": 0, "unit_11": 0, "unit_8": 0, "unit_0": 0, "unit_3": 0, "unit_7": 0, "unit_10": 0, "unit_5": 0, "unit_4": 0, "unit_9": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:2", "state_id": "catalog_lookup_v2:train:2", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00198; color=pink; size=small; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "pink, small", "candidate_1": "yellow, medium", "candidate_2": "blue, large", "candidate_3": "cyan, medium", "candidate_4": "black, large", "candidate_5": "pink, large", "candidate_6": "yellow, small", "candidate_7": "red, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00198\",\"pink\",\"small\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_4": 0, "candidate_0": 1, "candidate_6": 0, "candidate_2": 0, "candidate_1": 0, "candidate_3": 0, "candidate_5": 0, "candidate_7": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:3", "state_id": "catalog_lookup_v2:train:3", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00050; color=green; size=small; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "black, small", "candidate_1": "red, large", "candidate_2": "yellow, small", "candidate_3": "black, medium", "candidate_4": "red, medium", "candidate_5": "blue, medium", "candidate_6": "yellow, medium", "candidate_7": "pink, medium", "candidate_8": "blue, large", "candidate_9": "green, small", "candidate_10": "cyan, small", "candidate_11": "yellow, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_9", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 1, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00050\",\"green\",\"small\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_3": 0, "candidate_2": 0, "candidate_11": 0, "candidate_4": 0, "candidate_5": 0, "candidate_9": 1, "candidate_8": 0, "candidate_0": 0, "candidate_6": 0, "candidate_1": 0, "candidate_10": 0, "candidate_7": 0}, "ready": {"false": 0.91, "true": 0.09}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:train:197", "state_id": "smart_home_v2:train:197", "family_id": "smart_home_v2", "split": "train", "state": "Request: on the fan in the attic. Authorization=yes; people present=3; time=11:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom heater", "unit_1": "library speaker", "unit_2": "attic fan", "unit_3": "kitchen fan", "unit_4": "garden lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1, "unit_3": 0, "unit_4": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"attic\",\"fan\",true,3]", "template_id": "smart_home_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_2": 1, "unit_1": 0, "unit_4": 0, "unit_3": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:5", "state_id": "catalog_lookup_v2:train:5", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00192; color=pink; size=small; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "blue, large", "candidate_1": "pink, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00192\",\"pink\",\"small\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 1, "candidate_0": 0}, "ready": {"false": 0.76, "true": 0.24}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:4", "state_id": "catalog_lookup_v2:train:4", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00102; color=purple; size=small; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "white, medium", "candidate_1": "green, large", "candidate_2": "black, medium", "candidate_3": "purple, large", "candidate_4": "yellow, large", "candidate_5": "cyan, small", "candidate_6": "orange, large", "candidate_7": "red, large", "candidate_8": "pink, large", "candidate_9": "blue, large", "candidate_10": "orange, small", "candidate_11": "purple, small", "candidate_12": "blue, small", "candidate_13": "gray, large", "candidate_14": "black, small", "candidate_15": "pink, medium", "candidate_16": "gray, small", "candidate_17": "cyan, large", "candidate_18": "brown, large", "candidate_19": "brown, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_11", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 1, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00102\",\"purple\",\"small\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_14": 0, "candidate_18": 0, "candidate_9": 0, "candidate_4": 0, "candidate_1": 0, "candidate_7": 0, "candidate_19": 0, "candidate_5": 0, "candidate_17": 0, "candidate_0": 0, "candidate_2": 0, "candidate_10": 0, "candidate_12": 0, "candidate_8": 0, "candidate_13": 0, "candidate_11": 1, "candidate_15": 0, "candidate_6": 0, "candidate_16": 0, "candidate_3": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:6", "state_id": "catalog_lookup_v2:train:6", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00079; color=yellow; size=small; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "brown, medium", "candidate_1": "blue, small", "candidate_2": "white, large", "candidate_3": "purple, medium", "candidate_4": "yellow, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_4", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00079\",\"yellow\",\"small\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_4": 1, "candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:9", "state_id": "catalog_lookup_v2:train:9", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00184; color=orange; size=large; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, medium", "candidate_1": "cyan, large", "candidate_2": "brown, medium", "candidate_3": "white, small", "candidate_4": "brown, small", "candidate_5": "brown, large", "candidate_6": "gray, large", "candidate_7": "cyan, small", "candidate_8": "white, medium", "candidate_9": "orange, large", "candidate_10": "gray, small", "candidate_11": "green, large", "candidate_12": "purple, small", "candidate_13": "yellow, medium", "candidate_14": "pink, medium", "candidate_15": "white, large", "candidate_16": "purple, large", "candidate_17": "red, medium", "candidate_18": "yellow, small", "candidate_19": "orange, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_9", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 1, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00184\",\"orange\",\"large\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_14": 0, "candidate_18": 0, "candidate_9": 1, "candidate_4": 0, "candidate_1": 0, "candidate_17": 0, "candidate_5": 0, "candidate_19": 0, "candidate_7": 0, "candidate_0": 0, "candidate_2": 0, "candidate_10": 0, "candidate_15": 0, "candidate_8": 0, "candidate_13": 0, "candidate_11": 0, "candidate_12": 0, "candidate_3": 0, "candidate_6": 0, "candidate_16": 0}, "ready": {"false": 0.77, "true": 0.23}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:10", "state_id": "catalog_lookup_v2:train:10", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00222; color=brown; size=small; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "cyan, large", "candidate_1": "brown, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00222\",\"brown\",\"small\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 1, "candidate_0": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:7", "state_id": "catalog_lookup_v2:train:7", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00199; color=pink; size=small; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, large", "candidate_1": "brown, small", "candidate_2": "yellow, medium", "candidate_3": "pink, small", "candidate_4": "white, small", "candidate_5": "blue, large", "candidate_6": "brown, medium", "candidate_7": "green, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_3", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 1, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00199\",\"pink\",\"small\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_4": 0, "candidate_1": 0, "candidate_6": 0, "candidate_2": 0, "candidate_0": 0, "candidate_5": 0, "candidate_7": 0, "candidate_3": 1}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:8", "state_id": "catalog_lookup_v2:train:8", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00221; color=brown; size=small; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "orange, large", "candidate_1": "green, small", "candidate_2": "brown, small", "candidate_3": "white, large", "candidate_4": "green, large", "candidate_5": "orange, small", "candidate_6": "blue, large", "candidate_7": "blue, small", "candidate_8": "cyan, medium", "candidate_9": "pink, small", "candidate_10": "black, medium", "candidate_11": "yellow, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_2", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 1, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00221\",\"brown\",\"small\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_7": 0, "candidate_5": 0, "candidate_3": 0, "candidate_8": 0, "candidate_10": 0, "candidate_9": 0, "candidate_4": 0, "candidate_6": 0, "candidate_11": 0, "candidate_2": 1, "candidate_0": 0, "candidate_1": 0}, "ready": {"false": 0.16000000000000003, "true": 0.84}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:11", "state_id": "catalog_lookup_v2:train:11", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00045; color=blue; size=large; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, small", "candidate_1": "gray, small", "candidate_2": "brown, medium", "candidate_3": "blue, medium", "candidate_4": "blue, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_4", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 1}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00045\",\"blue\",\"large\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_0": 0, "candidate_4": 1, "candidate_2": 0, "candidate_3": 0}, "ready": {"false": 0.13, "true": 0.87}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:16", "state_id": "catalog_lookup_v2:train:16", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00021; color=red; size=large; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "brown, medium", "candidate_1": "red, large", "candidate_2": "purple, medium", "candidate_3": "green, medium", "candidate_4": "blue, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00021\",\"red\",\"large\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_0": 0, "candidate_4": 0}, "ready": {"false": 0.13, "true": 0.87}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:15", "state_id": "catalog_lookup_v2:train:15", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00071; color=green; size=large; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "black, medium", "candidate_1": "green, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00071\",\"green\",\"large\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 1, "candidate_0": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:12", "state_id": "catalog_lookup_v2:train:12", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00259; color=gray; size=large; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "brown, large", "candidate_1": "green, small", "candidate_2": "gray, large", "candidate_3": "yellow, medium", "candidate_4": "purple, medium", "candidate_5": "blue, medium", "candidate_6": "black, small", "candidate_7": "purple, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_2", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 1, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00259\",\"gray\",\"large\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_5": 0, "candidate_4": 0, "candidate_7": 0, "candidate_1": 0, "candidate_0": 0, "candidate_3": 0, "candidate_6": 0, "candidate_2": 1}, "ready": {"false": 0.9299999999999999, "true": 0.07}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:14", "state_id": "catalog_lookup_v2:train:14", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00212; color=pink; size=large; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "pink, medium", "candidate_1": "blue, large", "candidate_2": "black, small", "candidate_3": "black, medium", "candidate_4": "brown, small", "candidate_5": "purple, medium", "candidate_6": "brown, medium", "candidate_7": "green, large", "candidate_8": "yellow, small", "candidate_9": "blue, medium", "candidate_10": "white, medium", "candidate_11": "brown, large", "candidate_12": "white, large", "candidate_13": "cyan, small", "candidate_14": "yellow, medium", "candidate_15": "green, medium", "candidate_16": "gray, medium", "candidate_17": "red, medium", "candidate_18": "pink, large", "candidate_19": "cyan, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_18", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 1, "candidate_19": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00212\",\"pink\",\"large\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_18": 1, "candidate_5": 0, "candidate_12": 0, "candidate_15": 0, "candidate_7": 0, "candidate_4": 0, "candidate_3": 0, "candidate_6": 0, "candidate_14": 0, "candidate_2": 0, "candidate_17": 0, "candidate_11": 0, "candidate_13": 0, "candidate_1": 0, "candidate_16": 0, "candidate_0": 0, "candidate_10": 0, "candidate_8": 0, "candidate_19": 0, "candidate_9": 0}, "ready": {"false": 0.17000000000000004, "true": 0.83}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:13", "state_id": "catalog_lookup_v2:train:13", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00268; color=cyan; size=small; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "orange, medium", "candidate_1": "black, small", "candidate_2": "gray, small", "candidate_3": "purple, small", "candidate_4": "black, medium", "candidate_5": "green, medium", "candidate_6": "green, large", "candidate_7": "white, large", "candidate_8": "brown, medium", "candidate_9": "brown, large", "candidate_10": "cyan, small", "candidate_11": "blue, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_10", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 1, "candidate_11": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00268\",\"cyan\",\"small\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_9": 0, "candidate_5": 0, "candidate_6": 0, "candidate_3": 0, "candidate_8": 0, "candidate_0": 0, "candidate_2": 0, "candidate_4": 0, "candidate_10": 1, "candidate_7": 0, "candidate_1": 0, "candidate_11": 0}, "ready": {"false": 0.15000000000000002, "true": 0.85}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:17", "state_id": "catalog_lookup_v2:train:17", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00025; color=blue; size=small; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, small", "candidate_1": "blue, small", "candidate_2": "brown, small", "candidate_3": "gray, large", "candidate_4": "red, medium", "candidate_5": "yellow, small", "candidate_6": "brown, medium", "candidate_7": "pink, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00025\",\"blue\",\"small\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_4": 0, "candidate_3": 0, "candidate_7": 0, "candidate_1": 1, "candidate_5": 0, "candidate_6": 0, "candidate_0": 0, "candidate_2": 0}, "ready": {"false": 0.81, "true": 0.19}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:18", "state_id": "catalog_lookup_v2:train:18", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00279; color=cyan; size=medium; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "green, medium", "candidate_1": "black, small", "candidate_2": "brown, medium", "candidate_3": "gray, medium", "candidate_4": "cyan, medium", "candidate_5": "blue, large", "candidate_6": "brown, small", "candidate_7": "white, medium", "candidate_8": "red, medium", "candidate_9": "yellow, small", "candidate_10": "pink, small", "candidate_11": "pink, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_4", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 1, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00279\",\"cyan\",\"medium\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_4": 1, "candidate_0": 0, "candidate_9": 0, "candidate_10": 0, "candidate_6": 0, "candidate_11": 0, "candidate_2": 0, "candidate_7": 0, "candidate_3": 0, "candidate_5": 0, "candidate_1": 0, "candidate_8": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:19", "state_id": "catalog_lookup_v2:train:19", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00153; color=black; size=medium; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, large", "candidate_1": "orange, small", "candidate_2": "gray, large", "candidate_3": "red, medium", "candidate_4": "pink, medium", "candidate_5": "red, large", "candidate_6": "green, small", "candidate_7": "purple, small", "candidate_8": "orange, medium", "candidate_9": "blue, large", "candidate_10": "yellow, medium", "candidate_11": "black, medium", "candidate_12": "black, large", "candidate_13": "gray, medium", "candidate_14": "white, medium", "candidate_15": "black, small", "candidate_16": "gray, small", "candidate_17": "cyan, medium", "candidate_18": "white, small", "candidate_19": "brown, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_11", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 1, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00153\",\"black\",\"medium\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_5": 0, "candidate_2": 0, "candidate_3": 0, "candidate_10": 0, "candidate_14": 0, "candidate_1": 0, "candidate_16": 0, "candidate_4": 0, "candidate_19": 0, "candidate_8": 0, "candidate_9": 0, "candidate_17": 0, "candidate_18": 0, "candidate_0": 0, "candidate_12": 0, "candidate_6": 0, "candidate_15": 0, "candidate_11": 1, "candidate_13": 0, "candidate_7": 0}, "ready": {"false": 0.79, "true": 0.21}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:20", "state_id": "catalog_lookup_v2:train:20", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00224; color=brown; size=medium; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "pink, medium", "candidate_1": "brown, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00224\",\"brown\",\"medium\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 0.77, "true": 0.23}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:21", "state_id": "catalog_lookup_v2:train:21", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00204; color=pink; size=medium; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "white, medium", "candidate_1": "pink, medium", "candidate_2": "gray, large", "candidate_3": "yellow, small", "candidate_4": "blue, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00204\",\"pink\",\"medium\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_2": 0, "candidate_4": 0, "candidate_0": 0, "candidate_3": 0, "candidate_1": 1}, "ready": {"false": 0.16000000000000003, "true": 0.84}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:23", "state_id": "catalog_lookup_v2:train:23", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00141; color=white; size=large; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "white, small", "candidate_1": "purple, medium", "candidate_2": "purple, small", "candidate_3": "yellow, small", "candidate_4": "black, large", "candidate_5": "red, medium", "candidate_6": "pink, small", "candidate_7": "black, medium", "candidate_8": "black, small", "candidate_9": "white, large", "candidate_10": "yellow, medium", "candidate_11": "green, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_9", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 1, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00141\",\"white\",\"large\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_8": 0, "candidate_4": 0, "candidate_3": 0, "candidate_9": 1, "candidate_10": 0, "candidate_5": 0, "candidate_1": 0, "candidate_7": 0, "candidate_6": 0, "candidate_11": 0, "candidate_0": 0, "candidate_2": 0}, "ready": {"false": 0.13, "true": 0.87}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:24", "state_id": "catalog_lookup_v2:train:24", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00194; color=pink; size=small; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, medium", "candidate_1": "pink, small", "candidate_2": "red, medium", "candidate_3": "black, large", "candidate_4": "cyan, medium", "candidate_5": "yellow, large", "candidate_6": "brown, small", "candidate_7": "yellow, medium", "candidate_8": "white, small", "candidate_9": "cyan, large", "candidate_10": "gray, small", "candidate_11": "orange, large", "candidate_12": "blue, small", "candidate_13": "gray, medium", "candidate_14": "white, medium", "candidate_15": "purple, small", "candidate_16": "pink, large", "candidate_17": "red, small", "candidate_18": "green, large", "candidate_19": "gray, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00194\",\"pink\",\"small\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_5": 0, "candidate_14": 0, "candidate_8": 0, "candidate_2": 0, "candidate_6": 0, "candidate_7": 0, "candidate_12": 0, "candidate_13": 0, "candidate_15": 0, "candidate_9": 0, "candidate_3": 0, "candidate_18": 0, "candidate_17": 0, "candidate_4": 0, "candidate_19": 0, "candidate_11": 0, "candidate_0": 0, "candidate_16": 0, "candidate_10": 0, "candidate_1": 1}, "ready": {"false": 0.91, "true": 0.09}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:22", "state_id": "catalog_lookup_v2:train:22", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00084; color=yellow; size=medium; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "yellow, medium", "candidate_1": "green, large", "candidate_2": "brown, large", "candidate_3": "purple, small", "candidate_4": "cyan, large", "candidate_5": "gray, large", "candidate_6": "orange, small", "candidate_7": "green, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00084\",\"yellow\",\"medium\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_5": 0, "candidate_2": 0, "candidate_0": 1, "candidate_3": 0, "candidate_1": 0, "candidate_6": 0, "candidate_4": 0, "candidate_7": 0}, "ready": {"false": 0.17000000000000004, "true": 0.83}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:26", "state_id": "catalog_lookup_v2:train:26", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00193; color=pink; size=small; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "black, medium", "candidate_1": "green, medium", "candidate_2": "blue, small", "candidate_3": "pink, small", "candidate_4": "green, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_3", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 1, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00193\",\"pink\",\"small\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_4": 0, "candidate_3": 1, "candidate_0": 0, "candidate_2": 0}, "ready": {"false": 0.79, "true": 0.21}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:25", "state_id": "catalog_lookup_v2:train:25", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00261; color=gray; size=large; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "gray, large", "candidate_1": "black, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00261\",\"gray\",\"large\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 0.14, "true": 0.86}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:29", "state_id": "catalog_lookup_v2:train:29", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00149; color=black; size=small; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "orange, medium", "candidate_1": "yellow, small", "candidate_2": "white, large", "candidate_3": "black, medium", "candidate_4": "cyan, small", "candidate_5": "black, small", "candidate_6": "brown, large", "candidate_7": "purple, large", "candidate_8": "yellow, large", "candidate_9": "cyan, large", "candidate_10": "white, small", "candidate_11": "blue, small", "candidate_12": "orange, small", "candidate_13": "pink, large", "candidate_14": "brown, small", "candidate_15": "yellow, medium", "candidate_16": "gray, small", "candidate_17": "white, medium", "candidate_18": "black, large", "candidate_19": "blue, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_5", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 1, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00149\",\"black\",\"small\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_13": 0, "candidate_6": 0, "candidate_1": 0, "candidate_17": 0, "candidate_19": 0, "candidate_5": 1, "candidate_14": 0, "candidate_4": 0, "candidate_0": 0, "candidate_8": 0, "candidate_15": 0, "candidate_3": 0, "candidate_18": 0, "candidate_9": 0, "candidate_16": 0, "candidate_12": 0, "candidate_10": 0, "candidate_11": 0, "candidate_7": 0, "candidate_2": 0}, "ready": {"false": 0.13, "true": 0.87}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:28", "state_id": "catalog_lookup_v2:train:28", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00252; color=gray; size=medium; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "cyan, large", "candidate_1": "orange, large", "candidate_2": "gray, medium", "candidate_3": "pink, small", "candidate_4": "gray, small", "candidate_5": "gray, large", "candidate_6": "white, large", "candidate_7": "red, large", "candidate_8": "black, small", "candidate_9": "blue, large", "candidate_10": "green, small", "candidate_11": "blue, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_2", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 1, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00252\",\"gray\",\"medium\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_11": 0, "candidate_10": 0, "candidate_5": 0, "candidate_2": 1, "candidate_8": 0, "candidate_6": 0, "candidate_4": 0, "candidate_0": 0, "candidate_7": 0, "candidate_3": 0, "candidate_9": 0}, "ready": {"false": 0.15000000000000002, "true": 0.85}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:30", "state_id": "catalog_lookup_v2:train:30", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00216; color=brown; size=small; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "yellow, large", "candidate_1": "brown, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00216\",\"brown\",\"small\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 0.76, "true": 0.24}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:33", "state_id": "catalog_lookup_v2:train:33", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00033; color=blue; size=medium; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "orange, small", "candidate_1": "yellow, medium", "candidate_2": "orange, large", "candidate_3": "white, large", "candidate_4": "yellow, small", "candidate_5": "gray, large", "candidate_6": "blue, large", "candidate_7": "brown, medium", "candidate_8": "blue, medium", "candidate_9": "green, small", "candidate_10": "brown, small", "candidate_11": "cyan, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_8", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 1, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00033\",\"blue\",\"medium\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_4": 0, "candidate_1": 0, "candidate_6": 0, "candidate_9": 0, "candidate_0": 0, "candidate_8": 1, "candidate_2": 0, "candidate_7": 0, "candidate_10": 0, "candidate_11": 0, "candidate_3": 0, "candidate_5": 0}, "ready": {"false": 0.8, "true": 0.2}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:31", "state_id": "catalog_lookup_v2:train:31", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00077; color=yellow; size=small; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "orange, small", "candidate_1": "yellow, small", "candidate_2": "blue, large", "candidate_3": "white, medium", "candidate_4": "gray, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00077\",\"yellow\",\"small\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_2": 0, "candidate_4": 0, "candidate_0": 0, "candidate_3": 0, "candidate_1": 1}, "ready": {"false": 0.13, "true": 0.87}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:27", "state_id": "catalog_lookup_v2:train:27", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00115; color=purple; size=large; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "red, small", "candidate_1": "black, large", "candidate_2": "red, medium", "candidate_3": "brown, large", "candidate_4": "pink, medium", "candidate_5": "cyan, medium", "candidate_6": "green, medium", "candidate_7": "purple, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_7", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00115\",\"purple\",\"large\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_5": 0, "candidate_6": 0, "candidate_3": 0, "candidate_7": 1, "candidate_0": 0, "candidate_2": 0, "candidate_4": 0, "candidate_1": 0}, "ready": {"false": 0.95, "true": 0.05}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:32", "state_id": "catalog_lookup_v2:train:32", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00104; color=purple; size=medium; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "yellow, large", "candidate_1": "cyan, large", "candidate_2": "orange, small", "candidate_3": "pink, medium", "candidate_4": "cyan, small", "candidate_5": "brown, medium", "candidate_6": "purple, medium", "candidate_7": "green, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_6", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 1, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00104\",\"purple\",\"medium\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_3": 0, "candidate_7": 0, "candidate_2": 0, "candidate_0": 0, "candidate_5": 0, "candidate_6": 1, "candidate_4": 0, "candidate_1": 0}, "ready": {"false": 0.79, "true": 0.21}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:34", "state_id": "catalog_lookup_v2:train:34", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00139; color=white; size=large; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "white, medium", "candidate_1": "brown, medium", "candidate_2": "green, large", "candidate_3": "pink, small", "candidate_4": "green, medium", "candidate_5": "pink, medium", "candidate_6": "black, small", "candidate_7": "gray, large", "candidate_8": "brown, large", "candidate_9": "red, small", "candidate_10": "cyan, large", "candidate_11": "yellow, small", "candidate_12": "orange, small", "candidate_13": "purple, medium", "candidate_14": "white, large", "candidate_15": "blue, medium", "candidate_16": "blue, small", "candidate_17": "gray, medium", "candidate_18": "cyan, medium", "candidate_19": "purple, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_14", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 1, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00139\",\"white\",\"large\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_13": 0, "candidate_1": 0, "candidate_14": 1, "candidate_17": 0, "candidate_5": 0, "candidate_6": 0, "candidate_19": 0, "candidate_4": 0, "candidate_0": 0, "candidate_8": 0, "candidate_15": 0, "candidate_3": 0, "candidate_18": 0, "candidate_12": 0, "candidate_16": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_7": 0, "candidate_2": 0}, "ready": {"false": 0.92, "true": 0.08}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:36", "state_id": "catalog_lookup_v2:train:36", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00169; color=orange; size=small; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "gray, small", "candidate_1": "purple, medium", "candidate_2": "red, medium", "candidate_3": "orange, large", "candidate_4": "orange, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_4", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00169\",\"orange\",\"small\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_3": 0, "candidate_4": 1, "candidate_1": 0, "candidate_2": 0, "candidate_0": 0}, "ready": {"false": 0.79, "true": 0.21}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:35", "state_id": "catalog_lookup_v2:train:35", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00285; color=cyan; size=large; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "cyan, large", "candidate_1": "blue, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00285\",\"cyan\",\"large\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_0": 1}, "ready": {"false": 0.13, "true": 0.87}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:37", "state_id": "catalog_lookup_v2:train:37", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00172; color=orange; size=small; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "yellow, large", "candidate_1": "green, medium", "candidate_2": "pink, medium", "candidate_3": "gray, medium", "candidate_4": "white, medium", "candidate_5": "white, large", "candidate_6": "green, small", "candidate_7": "orange, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_7", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 1}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00172\",\"orange\",\"small\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_5": 0, "candidate_4": 0, "candidate_7": 1, "candidate_1": 0, "candidate_0": 0, "candidate_3": 0, "candidate_6": 0, "candidate_2": 0}, "ready": {"false": 0.16000000000000003, "true": 0.84}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:39", "state_id": "catalog_lookup_v2:train:39", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00016; color=red; size=large; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "red, large", "candidate_1": "yellow, large", "candidate_2": "brown, large", "candidate_3": "orange, medium", "candidate_4": "red, medium", "candidate_5": "blue, small", "candidate_6": "gray, medium", "candidate_7": "red, small", "candidate_8": "cyan, small", "candidate_9": "green, medium", "candidate_10": "purple, small", "candidate_11": "black, small", "candidate_12": "cyan, medium", "candidate_13": "purple, large", "candidate_14": "pink, medium", "candidate_15": "white, large", "candidate_16": "cyan, large", "candidate_17": "brown, medium", "candidate_18": "gray, large", "candidate_19": "green, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00016\",\"red\",\"large\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_9": 0, "candidate_2": 0, "candidate_17": 0, "candidate_16": 0, "candidate_0": 1, "candidate_4": 0, "candidate_13": 0, "candidate_10": 0, "candidate_19": 0, "candidate_3": 0, "candidate_11": 0, "candidate_15": 0, "candidate_14": 0, "candidate_7": 0, "candidate_1": 0, "candidate_12": 0, "candidate_6": 0, "candidate_8": 0, "candidate_5": 0, "candidate_18": 0}, "ready": {"false": 0.78, "true": 0.22}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:38", "state_id": "catalog_lookup_v2:train:38", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00067; color=green; size=large; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "blue, small", "candidate_1": "red, large", "candidate_2": "pink, small", "candidate_3": "white, medium", "candidate_4": "gray, large", "candidate_5": "cyan, medium", "candidate_6": "purple, medium", "candidate_7": "gray, medium", "candidate_8": "pink, medium", "candidate_9": "green, large", "candidate_10": "green, medium", "candidate_11": "green, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_9", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 1, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00067\",\"green\",\"large\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_8": 0, "candidate_4": 0, "candidate_9": 1, "candidate_5": 0, "candidate_10": 0, "candidate_2": 0, "candidate_3": 0, "candidate_7": 0, "candidate_11": 0, "candidate_0": 0, "candidate_1": 0, "candidate_6": 0}, "ready": {"false": 0.92, "true": 0.08}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:42", "state_id": "catalog_lookup_v2:train:42", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00035; color=blue; size=medium; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, large", "candidate_1": "red, medium", "candidate_2": "blue, medium", "candidate_3": "cyan, small", "candidate_4": "gray, medium", "candidate_5": "green, medium", "candidate_6": "white, small", "candidate_7": "cyan, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_2", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 1, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00035\",\"blue\",\"medium\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_4": 0, "candidate_5": 0, "candidate_2": 1, "candidate_3": 0, "candidate_7": 0, "candidate_0": 0, "candidate_1": 0, "candidate_6": 0}, "ready": {"false": 0.94, "true": 0.06}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:43", "state_id": "catalog_lookup_v2:train:43", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00142; color=white; size=large; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "white, large", "candidate_1": "red, medium", "candidate_2": "cyan, medium", "candidate_3": "green, medium", "candidate_4": "red, small", "candidate_5": "gray, small", "candidate_6": "yellow, small", "candidate_7": "black, medium", "candidate_8": "white, small", "candidate_9": "brown, small", "candidate_10": "blue, large", "candidate_11": "orange, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00142\",\"white\",\"large\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_3": 0, "candidate_6": 0, "candidate_7": 0, "candidate_2": 0, "candidate_8": 0, "candidate_0": 1, "candidate_5": 0, "candidate_11": 0, "candidate_4": 0, "candidate_1": 0, "candidate_9": 0, "candidate_10": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:41", "state_id": "catalog_lookup_v2:train:41", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00148; color=black; size=small; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "orange, small", "candidate_1": "green, medium", "candidate_2": "black, medium", "candidate_3": "blue, medium", "candidate_4": "black, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_4", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 1}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00148\",\"black\",\"small\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_3": 0, "candidate_0": 0, "candidate_2": 0, "candidate_4": 1, "candidate_1": 0}, "ready": {"false": 0.18000000000000005, "true": 0.82}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:40", "state_id": "catalog_lookup_v2:train:40", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00068; color=green; size=large; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, small", "candidate_1": "green, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00068\",\"green\",\"large\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 1, "candidate_0": 0}, "ready": {"false": 0.15000000000000002, "true": 0.85}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:44", "state_id": "catalog_lookup_v2:train:44", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00039; color=blue; size=medium; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "gray, small", "candidate_1": "pink, small", "candidate_2": "brown, medium", "candidate_3": "green, small", "candidate_4": "purple, small", "candidate_5": "brown, large", "candidate_6": "blue, large", "candidate_7": "green, medium", "candidate_8": "cyan, small", "candidate_9": "red, medium", "candidate_10": "blue, medium", "candidate_11": "yellow, large", "candidate_12": "cyan, large", "candidate_13": "red, small", "candidate_14": "black, medium", "candidate_15": "green, large", "candidate_16": "orange, small", "candidate_17": "brown, small", "candidate_18": "gray, large", "candidate_19": "blue, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_10", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 1, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00039\",\"blue\",\"medium\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_16": 0, "candidate_2": 0, "candidate_3": 0, "candidate_10": 1, "candidate_14": 0, "candidate_1": 0, "candidate_19": 0, "candidate_4": 0, "candidate_5": 0, "candidate_9": 0, "candidate_8": 0, "candidate_17": 0, "candidate_18": 0, "candidate_0": 0, "candidate_12": 0, "candidate_6": 0, "candidate_15": 0, "candidate_11": 0, "candidate_13": 0, "candidate_7": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:46", "state_id": "catalog_lookup_v2:train:46", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00271; color=cyan; size=small; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "yellow, large", "candidate_1": "brown, large", "candidate_2": "cyan, small", "candidate_3": "red, medium", "candidate_4": "orange, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_2", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 1, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00271\",\"cyan\",\"small\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_4": 0, "candidate_3": 0, "candidate_2": 1, "candidate_0": 0, "candidate_1": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:45", "state_id": "catalog_lookup_v2:train:45", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00128; color=white; size=medium; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "green, large", "candidate_1": "white, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00128\",\"white\",\"medium\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 0.76, "true": 0.24}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:51", "state_id": "catalog_lookup_v2:train:51", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00082; color=yellow; size=medium; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, medium", "candidate_1": "yellow, medium", "candidate_2": "red, small", "candidate_3": "blue, large", "candidate_4": "red, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00082\",\"yellow\",\"medium\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 1, "candidate_4": 0, "candidate_3": 0, "candidate_0": 0, "candidate_2": 0}, "ready": {"false": 0.94, "true": 0.06}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:47", "state_id": "catalog_lookup_v2:train:47", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00047; color=blue; size=large; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "white, medium", "candidate_1": "white, small", "candidate_2": "brown, medium", "candidate_3": "black, medium", "candidate_4": "orange, large", "candidate_5": "blue, large", "candidate_6": "green, large", "candidate_7": "pink, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_5", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 1, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00047\",\"blue\",\"large\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_5": 1, "candidate_1": 0, "candidate_2": 0, "candidate_6": 0, "candidate_0": 0, "candidate_7": 0, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:50", "state_id": "catalog_lookup_v2:train:50", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00042; color=blue; size=large; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "blue, large", "candidate_1": "gray, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00042\",\"blue\",\"large\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_0": 1}, "ready": {"false": 0.92, "true": 0.08}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:49", "state_id": "catalog_lookup_v2:train:49", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00185; color=orange; size=large; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "white, medium", "candidate_1": "yellow, small", "candidate_2": "brown, medium", "candidate_3": "gray, medium", "candidate_4": "red, large", "candidate_5": "gray, large", "candidate_6": "black, medium", "candidate_7": "orange, medium", "candidate_8": "blue, large", "candidate_9": "black, small", "candidate_10": "orange, large", "candidate_11": "pink, small", "candidate_12": "pink, medium", "candidate_13": "green, medium", "candidate_14": "purple, medium", "candidate_15": "yellow, large", "candidate_16": "green, small", "candidate_17": "brown, small", "candidate_18": "white, small", "candidate_19": "gray, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_10", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 1, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00185\",\"orange\",\"large\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_17": 0, "candidate_14": 0, "candidate_9": 0, "candidate_0": 0, "candidate_10": 1, "candidate_11": 0, "candidate_3": 0, "candidate_19": 0, "candidate_16": 0, "candidate_8": 0, "candidate_12": 0, "candidate_13": 0, "candidate_4": 0, "candidate_18": 0, "candidate_6": 0, "candidate_15": 0, "candidate_2": 0, "candidate_1": 0, "candidate_7": 0, "candidate_5": 0}, "ready": {"false": 0.8, "true": 0.2}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:52", "state_id": "catalog_lookup_v2:train:52", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00159; color=black; size=medium; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "blue, medium", "candidate_1": "purple, small", "candidate_2": "yellow, large", "candidate_3": "orange, small", "candidate_4": "red, medium", "candidate_5": "black, medium", "candidate_6": "blue, small", "candidate_7": "yellow, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_5", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 1, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00159\",\"black\",\"medium\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_3": 0, "candidate_1": 0, "candidate_6": 0, "candidate_5": 1, "candidate_0": 0, "candidate_4": 0, "candidate_7": 0, "candidate_2": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:48", "state_id": "catalog_lookup_v2:train:48", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00267; color=cyan; size=small; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "cyan, small", "candidate_1": "gray, large", "candidate_2": "yellow, large", "candidate_3": "orange, large", "candidate_4": "pink, small", "candidate_5": "brown, large", "candidate_6": "red, medium", "candidate_7": "gray, small", "candidate_8": "white, small", "candidate_9": "green, small", "candidate_10": "purple, large", "candidate_11": "black, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00267\",\"cyan\",\"small\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 1, "candidate_10": 0, "candidate_7": 0, "candidate_11": 0, "candidate_1": 0, "candidate_9": 0, "candidate_6": 0, "candidate_8": 0, "candidate_2": 0, "candidate_3": 0, "candidate_5": 0, "candidate_4": 0}, "ready": {"false": 0.94, "true": 0.06}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:54", "state_id": "catalog_lookup_v2:train:54", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00225; color=brown; size=medium; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "brown, medium", "candidate_1": "black, medium", "candidate_2": "orange, medium", "candidate_3": "blue, medium", "candidate_4": "cyan, large", "candidate_5": "yellow, large", "candidate_6": "purple, small", "candidate_7": "yellow, medium", "candidate_8": "cyan, medium", "candidate_9": "green, large", "candidate_10": "purple, medium", "candidate_11": "red, large", "candidate_12": "yellow, small", "candidate_13": "white, small", "candidate_14": "brown, small", "candidate_15": "gray, large", "candidate_16": "red, small", "candidate_17": "black, small", "candidate_18": "cyan, small", "candidate_19": "black, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00225\",\"brown\",\"medium\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_5": 0, "candidate_2": 0, "candidate_19": 0, "candidate_10": 0, "candidate_14": 0, "candidate_1": 0, "candidate_16": 0, "candidate_4": 0, "candidate_3": 0, "candidate_9": 0, "candidate_8": 0, "candidate_17": 0, "candidate_18": 0, "candidate_0": 1, "candidate_12": 0, "candidate_6": 0, "candidate_15": 0, "candidate_11": 0, "candidate_13": 0, "candidate_7": 0}, "ready": {"false": 0.8, "true": 0.2}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:53", "state_id": "catalog_lookup_v2:train:53", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00203; color=pink; size=medium; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "blue, large", "candidate_1": "brown, medium", "candidate_2": "red, large", "candidate_3": "green, medium", "candidate_4": "green, large", "candidate_5": "red, small", "candidate_6": "black, large", "candidate_7": "white, small", "candidate_8": "orange, medium", "candidate_9": "yellow, large", "candidate_10": "pink, medium", "candidate_11": "green, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_10", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 1, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00203\",\"pink\",\"medium\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_5": 0, "candidate_2": 0, "candidate_3": 0, "candidate_10": 1, "candidate_11": 0, "candidate_6": 0, "candidate_1": 0, "candidate_4": 0, "candidate_0": 0, "candidate_8": 0, "candidate_9": 0, "candidate_7": 0}, "ready": {"false": 0.94, "true": 0.06}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:55", "state_id": "catalog_lookup_v2:train:55", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00010; color=red; size=medium; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "green, medium", "candidate_1": "red, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00010\",\"red\",\"medium\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 1, "candidate_0": 0}, "ready": {"false": 0.9299999999999999, "true": 0.07}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:56", "state_id": "catalog_lookup_v2:train:56", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00205; color=pink; size=medium; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "pink, medium", "candidate_1": "purple, medium", "candidate_2": "red, medium", "candidate_3": "gray, large", "candidate_4": "blue, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00205\",\"pink\",\"medium\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 1, "candidate_1": 0, "candidate_4": 0, "candidate_3": 0, "candidate_2": 0}, "ready": {"false": 0.12, "true": 0.88}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:58", "state_id": "catalog_lookup_v2:train:58", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00180; color=orange; size=medium; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "gray, medium", "candidate_1": "gray, small", "candidate_2": "blue, large", "candidate_3": "red, large", "candidate_4": "yellow, small", "candidate_5": "orange, small", "candidate_6": "purple, medium", "candidate_7": "pink, large", "candidate_8": "green, large", "candidate_9": "green, small", "candidate_10": "orange, medium", "candidate_11": "white, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_10", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 1, "candidate_11": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00180\",\"orange\",\"medium\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_9": 0, "candidate_10": 1, "candidate_6": 0, "candidate_1": 0, "candidate_0": 0, "candidate_2": 0, "candidate_5": 0, "candidate_8": 0, "candidate_3": 0, "candidate_7": 0, "candidate_4": 0, "candidate_11": 0}, "ready": {"false": 0.18000000000000005, "true": 0.82}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:60", "state_id": "catalog_lookup_v2:train:60", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00136; color=white; size=large; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "white, large", "candidate_1": "purple, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00136\",\"white\",\"large\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 0.78, "true": 0.22}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:59", "state_id": "catalog_lookup_v2:train:59", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00093; color=yellow; size=large; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "brown, medium", "candidate_1": "yellow, large", "candidate_2": "blue, large", "candidate_3": "cyan, large", "candidate_4": "pink, medium", "candidate_5": "orange, medium", "candidate_6": "green, small", "candidate_7": "white, medium", "candidate_8": "black, small", "candidate_9": "green, large", "candidate_10": "purple, medium", "candidate_11": "purple, large", "candidate_12": "orange, small", "candidate_13": "white, small", "candidate_14": "orange, large", "candidate_15": "white, large", "candidate_16": "blue, small", "candidate_17": "gray, small", "candidate_18": "green, medium", "candidate_19": "black, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00093\",\"yellow\",\"large\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_18": 0, "candidate_16": 0, "candidate_2": 0, "candidate_0": 0, "candidate_19": 0, "candidate_1": 1, "candidate_8": 0, "candidate_13": 0, "candidate_6": 0, "candidate_7": 0, "candidate_15": 0, "candidate_9": 0, "candidate_3": 0, "candidate_14": 0, "candidate_10": 0, "candidate_4": 0, "candidate_11": 0, "candidate_17": 0, "candidate_5": 0, "candidate_12": 0}, "ready": {"false": 0.14, "true": 0.86}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:57", "state_id": "catalog_lookup_v2:train:57", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00073; color=yellow; size=small; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "black, large", "candidate_1": "red, medium", "candidate_2": "red, large", "candidate_3": "cyan, small", "candidate_4": "brown, small", "candidate_5": "cyan, medium", "candidate_6": "yellow, small", "candidate_7": "blue, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_6", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 1, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00073\",\"yellow\",\"small\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_2": 0, "candidate_5": 0, "candidate_6": 1, "candidate_4": 0, "candidate_0": 0, "candidate_7": 0, "candidate_3": 0}, "ready": {"false": 0.8, "true": 0.2}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:61", "state_id": "catalog_lookup_v2:train:61", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00240; color=gray; size=small; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "green, large", "candidate_1": "purple, small", "candidate_2": "gray, small", "candidate_3": "brown, large", "candidate_4": "orange, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_2", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 1, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00240\",\"gray\",\"small\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_4": 0, "candidate_3": 0, "candidate_0": 0, "candidate_2": 1, "candidate_1": 0}, "ready": {"false": 0.75, "true": 0.25}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:62", "state_id": "catalog_lookup_v2:train:62", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00281; color=cyan; size=large; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "brown, medium", "candidate_1": "red, medium", "candidate_2": "black, medium", "candidate_3": "white, small", "candidate_4": "blue, large", "candidate_5": "white, medium", "candidate_6": "cyan, large", "candidate_7": "orange, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_6", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 1, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00281\",\"cyan\",\"large\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_5": 0, "candidate_4": 0, "candidate_0": 0, "candidate_3": 0, "candidate_1": 0, "candidate_7": 0, "candidate_6": 1, "candidate_2": 0}, "ready": {"false": 0.8, "true": 0.2}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:63", "state_id": "catalog_lookup_v2:train:63", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00015; color=red; size=medium; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, large", "candidate_1": "yellow, large", "candidate_2": "yellow, small", "candidate_3": "brown, large", "candidate_4": "purple, small", "candidate_5": "cyan, large", "candidate_6": "blue, medium", "candidate_7": "red, medium", "candidate_8": "red, large", "candidate_9": "cyan, medium", "candidate_10": "gray, medium", "candidate_11": "green, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_7", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 1, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00015\",\"red\",\"medium\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_2": 0, "candidate_8": 0, "candidate_11": 0, "candidate_3": 0, "candidate_7": 1, "candidate_5": 0, "candidate_6": 0, "candidate_4": 0, "candidate_9": 0, "candidate_1": 0, "candidate_10": 0, "candidate_0": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:67", "state_id": "catalog_lookup_v2:train:67", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00123; color=white; size=small; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "red, large", "candidate_1": "purple, medium", "candidate_2": "gray, medium", "candidate_3": "pink, medium", "candidate_4": "gray, large", "candidate_5": "white, small", "candidate_6": "cyan, medium", "candidate_7": "black, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_5", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 1, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00123\",\"white\",\"small\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_7": 0, "candidate_5": 1, "candidate_3": 0, "candidate_6": 0, "candidate_4": 0, "candidate_2": 0, "candidate_0": 0, "candidate_1": 0}, "ready": {"false": 0.92, "true": 0.08}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:64", "state_id": "catalog_lookup_v2:train:64", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00076; color=yellow; size=small; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, large", "candidate_1": "black, large", "candidate_2": "cyan, small", "candidate_3": "black, medium", "candidate_4": "white, small", "candidate_5": "white, large", "candidate_6": "blue, medium", "candidate_7": "blue, large", "candidate_8": "purple, small", "candidate_9": "green, medium", "candidate_10": "blue, small", "candidate_11": "cyan, large", "candidate_12": "green, small", "candidate_13": "cyan, medium", "candidate_14": "brown, medium", "candidate_15": "pink, small", "candidate_16": "yellow, large", "candidate_17": "yellow, small", "candidate_18": "orange, medium", "candidate_19": "brown, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_17", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 1, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00076\",\"yellow\",\"small\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_4": 0, "candidate_16": 0, "candidate_0": 0, "candidate_9": 0, "candidate_7": 0, "candidate_18": 0, "candidate_19": 0, "candidate_15": 0, "candidate_5": 0, "candidate_8": 0, "candidate_10": 0, "candidate_6": 0, "candidate_14": 0, "candidate_1": 0, "candidate_17": 1, "candidate_12": 0, "candidate_2": 0, "candidate_11": 0, "candidate_3": 0, "candidate_13": 0}, "ready": {"false": 0.15000000000000002, "true": 0.85}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:66", "state_id": "catalog_lookup_v2:train:66", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00241; color=gray; size=small; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, medium", "candidate_1": "gray, small", "candidate_2": "brown, medium", "candidate_3": "cyan, medium", "candidate_4": "gray, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00241\",\"gray\",\"small\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_3": 0, "candidate_2": 0, "candidate_1": 1, "candidate_4": 0, "candidate_0": 0}, "ready": {"false": 0.79, "true": 0.21}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:65", "state_id": "catalog_lookup_v2:train:65", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00165; color=black; size=large; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "black, large", "candidate_1": "cyan, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00165\",\"black\",\"large\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_0": 1}, "ready": {"false": 0.12, "true": 0.88}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:69", "state_id": "catalog_lookup_v2:train:69", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00264; color=cyan; size=small; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "brown, medium", "candidate_1": "pink, large", "candidate_2": "brown, large", "candidate_3": "gray, large", "candidate_4": "red, medium", "candidate_5": "blue, medium", "candidate_6": "green, large", "candidate_7": "orange, medium", "candidate_8": "black, large", "candidate_9": "pink, medium", "candidate_10": "purple, medium", "candidate_11": "cyan, large", "candidate_12": "yellow, medium", "candidate_13": "cyan, medium", "candidate_14": "black, small", "candidate_15": "red, small", "candidate_16": "cyan, small", "candidate_17": "red, large", "candidate_18": "purple, large", "candidate_19": "pink, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_16", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 1, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00264\",\"cyan\",\"small\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_12": 0, "candidate_3": 0, "candidate_18": 0, "candidate_6": 0, "candidate_13": 0, "candidate_4": 0, "candidate_17": 0, "candidate_11": 0, "candidate_16": 1, "candidate_19": 0, "candidate_14": 0, "candidate_15": 0, "candidate_9": 0, "candidate_8": 0, "candidate_5": 0, "candidate_10": 0, "candidate_7": 0, "candidate_2": 0, "candidate_0": 0, "candidate_1": 0}, "ready": {"false": 0.76, "true": 0.24}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:70", "state_id": "catalog_lookup_v2:train:70", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00201; color=pink; size=medium; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "pink, medium", "candidate_1": "yellow, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00201\",\"pink\",\"medium\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_0": 1}, "ready": {"false": 0.79, "true": 0.21}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:68", "state_id": "catalog_lookup_v2:train:68", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00119; color=purple; size=large; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "blue, large", "candidate_1": "white, medium", "candidate_2": "white, large", "candidate_3": "orange, small", "candidate_4": "green, small", "candidate_5": "yellow, large", "candidate_6": "black, medium", "candidate_7": "purple, large", "candidate_8": "cyan, medium", "candidate_9": "pink, small", "candidate_10": "pink, medium", "candidate_11": "black, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_7", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 1, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00119\",\"purple\",\"large\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_5": 0, "candidate_2": 0, "candidate_0": 0, "candidate_10": 0, "candidate_11": 0, "candidate_6": 0, "candidate_1": 0, "candidate_4": 0, "candidate_3": 0, "candidate_9": 0, "candidate_8": 0, "candidate_7": 1}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:71", "state_id": "catalog_lookup_v2:train:71", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00103; color=purple; size=small; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "green, large", "candidate_1": "red, large", "candidate_2": "orange, large", "candidate_3": "purple, small", "candidate_4": "gray, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_3", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 1, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00103\",\"purple\",\"small\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_4": 0, "candidate_0": 0, "candidate_3": 1, "candidate_2": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:72", "state_id": "catalog_lookup_v2:train:72", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00233; color=brown; size=large; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "white, large", "candidate_1": "pink, small", "candidate_2": "cyan, medium", "candidate_3": "purple, medium", "candidate_4": "black, large", "candidate_5": "brown, large", "candidate_6": "cyan, small", "candidate_7": "purple, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_5", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 1, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00233\",\"brown\",\"large\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_5": 1, "candidate_6": 0, "candidate_7": 0, "candidate_0": 0, "candidate_3": 0, "candidate_2": 0, "candidate_4": 0, "candidate_1": 0}, "ready": {"false": 0.79, "true": 0.21}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:73", "state_id": "catalog_lookup_v2:train:73", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00270; color=cyan; size=small; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "brown, large", "candidate_1": "pink, medium", "candidate_2": "green, large", "candidate_3": "blue, large", "candidate_4": "red, large", "candidate_5": "red, medium", "candidate_6": "cyan, large", "candidate_7": "purple, medium", "candidate_8": "cyan, small", "candidate_9": "purple, large", "candidate_10": "green, medium", "candidate_11": "black, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_8", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 1, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00270\",\"cyan\",\"small\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_5": 0, "candidate_2": 0, "candidate_3": 0, "candidate_10": 0, "candidate_11": 0, "candidate_1": 0, "candidate_6": 0, "candidate_4": 0, "candidate_0": 0, "candidate_9": 0, "candidate_8": 1, "candidate_7": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:77", "state_id": "catalog_lookup_v2:train:77", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00223; color=brown; size=small; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, medium", "candidate_1": "brown, small", "candidate_2": "gray, medium", "candidate_3": "cyan, small", "candidate_4": "white, large", "candidate_5": "brown, medium", "candidate_6": "yellow, small", "candidate_7": "pink, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00223\",\"brown\",\"small\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_5": 0, "candidate_2": 0, "candidate_3": 0, "candidate_0": 0, "candidate_6": 0, "candidate_1": 1, "candidate_4": 0, "candidate_7": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:75", "state_id": "catalog_lookup_v2:train:75", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00092; color=yellow; size=large; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "yellow, large", "candidate_1": "white, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00092\",\"yellow\",\"large\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_0": 1}, "ready": {"false": 0.17000000000000004, "true": 0.83}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:74", "state_id": "catalog_lookup_v2:train:74", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00121; color=white; size=small; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "yellow, small", "candidate_1": "orange, medium", "candidate_2": "purple, medium", "candidate_3": "brown, small", "candidate_4": "gray, small", "candidate_5": "blue, large", "candidate_6": "pink, medium", "candidate_7": "black, small", "candidate_8": "yellow, large", "candidate_9": "pink, large", "candidate_10": "red, large", "candidate_11": "blue, medium", "candidate_12": "black, large", "candidate_13": "gray, medium", "candidate_14": "purple, large", "candidate_15": "black, medium", "candidate_16": "green, small", "candidate_17": "white, small", "candidate_18": "brown, medium", "candidate_19": "pink, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_17", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 1, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00121\",\"white\",\"small\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_18": 0, "candidate_5": 0, "candidate_12": 0, "candidate_15": 0, "candidate_7": 0, "candidate_4": 0, "candidate_3": 0, "candidate_6": 0, "candidate_14": 0, "candidate_13": 0, "candidate_17": 1, "candidate_11": 0, "candidate_2": 0, "candidate_1": 0, "candidate_16": 0, "candidate_0": 0, "candidate_10": 0, "candidate_8": 0, "candidate_19": 0, "candidate_9": 0}, "ready": {"false": 0.79, "true": 0.21}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:76", "state_id": "catalog_lookup_v2:train:76", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00000; color=red; size=small; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "gray, small", "candidate_1": "red, small", "candidate_2": "yellow, large", "candidate_3": "pink, large", "candidate_4": "purple, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00000\",\"red\",\"small\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 0, "candidate_2": 0, "candidate_4": 0, "candidate_1": 1, "candidate_3": 0}, "ready": {"false": 0.78, "true": 0.22}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:79", "state_id": "catalog_lookup_v2:train:79", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00061; color=green; size=medium; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "orange, medium", "candidate_1": "red, small", "candidate_2": "green, small", "candidate_3": "yellow, medium", "candidate_4": "pink, medium", "candidate_5": "pink, small", "candidate_6": "red, medium", "candidate_7": "black, small", "candidate_8": "cyan, medium", "candidate_9": "yellow, small", "candidate_10": "brown, medium", "candidate_11": "brown, small", "candidate_12": "green, large", "candidate_13": "white, large", "candidate_14": "red, large", "candidate_15": "white, medium", "candidate_16": "white, small", "candidate_17": "cyan, small", "candidate_18": "purple, large", "candidate_19": "green, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_19", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 1}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00061\",\"green\",\"medium\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_19": 1, "candidate_13": 0, "candidate_3": 0, "candidate_8": 0, "candidate_17": 0, "candidate_16": 0, "candidate_4": 0, "candidate_14": 0, "candidate_10": 0, "candidate_1": 0, "candidate_11": 0, "candidate_12": 0, "candidate_5": 0, "candidate_0": 0, "candidate_7": 0, "candidate_6": 0, "candidate_15": 0, "candidate_2": 0, "candidate_18": 0, "candidate_9": 0}, "ready": {"false": 0.13, "true": 0.87}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:78", "state_id": "catalog_lookup_v2:train:78", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00230; color=brown; size=medium; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "cyan, medium", "candidate_1": "red, small", "candidate_2": "yellow, small", "candidate_3": "purple, large", "candidate_4": "black, small", "candidate_5": "red, medium", "candidate_6": "black, large", "candidate_7": "gray, small", "candidate_8": "green, small", "candidate_9": "brown, small", "candidate_10": "white, small", "candidate_11": "brown, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_11", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00230\",\"brown\",\"medium\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_5": 0, "candidate_4": 0, "candidate_7": 0, "candidate_1": 0, "candidate_0": 0, "candidate_3": 0, "candidate_6": 0, "candidate_10": 0, "candidate_2": 0, "candidate_8": 0, "candidate_11": 1, "candidate_9": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:81", "state_id": "catalog_lookup_v2:train:81", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00066; color=green; size=large; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "gray, large", "candidate_1": "blue, large", "candidate_2": "green, large", "candidate_3": "yellow, small", "candidate_4": "brown, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_2", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 1, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00066\",\"green\",\"large\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_2": 1, "candidate_1": 0, "candidate_4": 0, "candidate_3": 0, "candidate_0": 0}, "ready": {"false": 0.91, "true": 0.09}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:84", "state_id": "catalog_lookup_v2:train:84", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00226; color=brown; size=medium; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "pink, medium", "candidate_1": "blue, small", "candidate_2": "black, medium", "candidate_3": "black, large", "candidate_4": "brown, large", "candidate_5": "yellow, large", "candidate_6": "gray, large", "candidate_7": "brown, medium", "candidate_8": "red, medium", "candidate_9": "red, large", "candidate_10": "purple, medium", "candidate_11": "white, medium", "candidate_12": "cyan, small", "candidate_13": "pink, large", "candidate_14": "cyan, large", "candidate_15": "white, small", "candidate_16": "red, small", "candidate_17": "gray, medium", "candidate_18": "green, large", "candidate_19": "orange, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_7", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 1, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00226\",\"brown\",\"medium\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_17": 0, "candidate_19": 0, "candidate_11": 0, "candidate_6": 0, "candidate_12": 0, "candidate_8": 0, "candidate_0": 0, "candidate_10": 0, "candidate_9": 0, "candidate_18": 0, "candidate_5": 0, "candidate_2": 0, "candidate_16": 0, "candidate_13": 0, "candidate_4": 0, "candidate_15": 0, "candidate_7": 1, "candidate_3": 0, "candidate_14": 0}, "ready": {"false": 0.92, "true": 0.08}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:82", "state_id": "catalog_lookup_v2:train:82", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00089; color=yellow; size=large; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "gray, small", "candidate_1": "yellow, large", "candidate_2": "red, small", "candidate_3": "gray, large", "candidate_4": "white, small", "candidate_5": "yellow, small", "candidate_6": "purple, small", "candidate_7": "pink, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00089\",\"yellow\",\"large\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_5": 0, "candidate_2": 0, "candidate_1": 1, "candidate_6": 0, "candidate_0": 0, "candidate_7": 0, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 0.81, "true": 0.19}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:80", "state_id": "catalog_lookup_v2:train:80", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00098; color=purple; size=small; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, small", "candidate_1": "yellow, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00098\",\"purple\",\"small\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_0": 1}, "ready": {"false": 0.92, "true": 0.08}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:85", "state_id": "catalog_lookup_v2:train:85", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00145; color=black; size=small; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "black, small", "candidate_1": "purple, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00145\",\"black\",\"small\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_0": 1}, "ready": {"false": 0.81, "true": 0.19}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:83", "state_id": "catalog_lookup_v2:train:83", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00002; color=red; size=small; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "yellow, medium", "candidate_1": "black, small", "candidate_2": "purple, small", "candidate_3": "purple, medium", "candidate_4": "green, large", "candidate_5": "brown, small", "candidate_6": "orange, medium", "candidate_7": "blue, large", "candidate_8": "black, large", "candidate_9": "pink, small", "candidate_10": "red, small", "candidate_11": "yellow, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_10", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 1, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00002\",\"red\",\"small\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_3": 0, "candidate_10": 1, "candidate_2": 0, "candidate_5": 0, "candidate_8": 0, "candidate_6": 0, "candidate_4": 0, "candidate_0": 0, "candidate_7": 0, "candidate_11": 0, "candidate_9": 0}, "ready": {"false": 0.9299999999999999, "true": 0.07}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:87", "state_id": "catalog_lookup_v2:train:87", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00027; color=blue; size=small; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, small", "candidate_1": "blue, small", "candidate_2": "blue, medium", "candidate_3": "gray, small", "candidate_4": "red, large", "candidate_5": "yellow, large", "candidate_6": "black, medium", "candidate_7": "cyan, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00027\",\"blue\",\"small\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_5": 0, "candidate_2": 0, "candidate_0": 0, "candidate_3": 0, "candidate_6": 0, "candidate_1": 1, "candidate_4": 0, "candidate_7": 0}, "ready": {"false": 0.9299999999999999, "true": 0.07}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:86", "state_id": "catalog_lookup_v2:train:86", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00004; color=red; size=small; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "red, small", "candidate_1": "gray, small", "candidate_2": "white, medium", "candidate_3": "purple, large", "candidate_4": "cyan, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00004\",\"red\",\"small\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 1, "candidate_4": 0, "candidate_1": 0, "candidate_3": 0, "candidate_2": 0}, "ready": {"false": 0.16000000000000003, "true": 0.84}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:88", "state_id": "catalog_lookup_v2:train:88", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00138; color=white; size=large; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "yellow, large", "candidate_1": "green, large", "candidate_2": "cyan, medium", "candidate_3": "black, large", "candidate_4": "purple, large", "candidate_5": "red, medium", "candidate_6": "white, large", "candidate_7": "gray, small", "candidate_8": "white, medium", "candidate_9": "blue, large", "candidate_10": "cyan, large", "candidate_11": "pink, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_6", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 1, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00138\",\"white\",\"large\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_5": 0, "candidate_0": 0, "candidate_7": 0, "candidate_8": 0, "candidate_3": 0, "candidate_2": 0, "candidate_4": 0, "candidate_10": 0, "candidate_6": 1, "candidate_9": 0, "candidate_11": 0}, "ready": {"false": 0.9299999999999999, "true": 0.07}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:90", "state_id": "catalog_lookup_v2:train:90", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00265; color=cyan; size=small; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "red, large", "candidate_1": "cyan, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00265\",\"cyan\",\"small\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 1, "candidate_0": 0}, "ready": {"false": 0.79, "true": 0.21}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:91", "state_id": "catalog_lookup_v2:train:91", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00096; color=purple; size=small; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "blue, small", "candidate_1": "gray, small", "candidate_2": "purple, small", "candidate_3": "orange, medium", "candidate_4": "cyan, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_2", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 1, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00096\",\"purple\",\"small\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_3": 0, "candidate_0": 0, "candidate_4": 0, "candidate_2": 1}, "ready": {"false": 0.77, "true": 0.23}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:89", "state_id": "catalog_lookup_v2:train:89", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00036; color=blue; size=medium; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, small", "candidate_1": "yellow, large", "candidate_2": "white, medium", "candidate_3": "purple, large", "candidate_4": "green, large", "candidate_5": "white, large", "candidate_6": "orange, large", "candidate_7": "cyan, small", "candidate_8": "pink, medium", "candidate_9": "purple, medium", "candidate_10": "black, small", "candidate_11": "cyan, large", "candidate_12": "gray, small", "candidate_13": "black, large", "candidate_14": "green, medium", "candidate_15": "blue, medium", "candidate_16": "brown, small", "candidate_17": "brown, large", "candidate_18": "red, medium", "candidate_19": "blue, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_15", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 1, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00036\",\"blue\",\"medium\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_16": 0, "candidate_19": 0, "candidate_18": 0, "candidate_12": 0, "candidate_6": 0, "candidate_1": 0, "candidate_13": 0, "candidate_14": 0, "candidate_17": 0, "candidate_9": 0, "candidate_8": 0, "candidate_7": 0, "candidate_11": 0, "candidate_4": 0, "candidate_0": 0, "candidate_2": 0, "candidate_10": 0, "candidate_15": 1, "candidate_5": 0, "candidate_3": 0}, "ready": {"false": 0.14, "true": 0.86}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:93", "state_id": "catalog_lookup_v2:train:93", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00107; color=purple; size=medium; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "gray, small", "candidate_1": "blue, medium", "candidate_2": "red, medium", "candidate_3": "purple, medium", "candidate_4": "brown, small", "candidate_5": "pink, medium", "candidate_6": "white, small", "candidate_7": "brown, medium", "candidate_8": "green, medium", "candidate_9": "orange, medium", "candidate_10": "red, small", "candidate_11": "cyan, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_3", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 1, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00107\",\"purple\",\"medium\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_5": 0, "candidate_2": 0, "candidate_0": 0, "candidate_10": 0, "candidate_11": 0, "candidate_6": 0, "candidate_3": 1, "candidate_4": 0, "candidate_1": 0, "candidate_9": 0, "candidate_8": 0, "candidate_7": 0}, "ready": {"false": 0.94, "true": 0.06}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:94", "state_id": "catalog_lookup_v2:train:94", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00250; color=gray; size=medium; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "brown, medium", "candidate_1": "blue, small", "candidate_2": "white, large", "candidate_3": "green, small", "candidate_4": "black, large", "candidate_5": "gray, small", "candidate_6": "green, large", "candidate_7": "yellow, large", "candidate_8": "blue, medium", "candidate_9": "orange, small", "candidate_10": "cyan, medium", "candidate_11": "orange, large", "candidate_12": "cyan, small", "candidate_13": "blue, large", "candidate_14": "red, large", "candidate_15": "gray, medium", "candidate_16": "black, medium", "candidate_17": "black, small", "candidate_18": "pink, small", "candidate_19": "orange, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_15", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 1, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00250\",\"gray\",\"medium\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_14": 0, "candidate_8": 0, "candidate_2": 0, "candidate_6": 0, "candidate_7": 0, "candidate_15": 1, "candidate_13": 0, "candidate_12": 0, "candidate_9": 0, "candidate_3": 0, "candidate_18": 0, "candidate_17": 0, "candidate_4": 0, "candidate_19": 0, "candidate_16": 0, "candidate_0": 0, "candidate_11": 0, "candidate_10": 0, "candidate_5": 0}, "ready": {"false": 0.9299999999999999, "true": 0.07}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:92", "state_id": "catalog_lookup_v2:train:92", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00022; color=red; size=large; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, large", "candidate_1": "red, small", "candidate_2": "black, large", "candidate_3": "red, large", "candidate_4": "green, large", "candidate_5": "gray, small", "candidate_6": "green, small", "candidate_7": "pink, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_3", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 1, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00022\",\"red\",\"large\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 0, "candidate_6": 0, "candidate_1": 0, "candidate_2": 0, "candidate_5": 0, "candidate_3": 1, "candidate_7": 0, "candidate_4": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:95", "state_id": "catalog_lookup_v2:train:95", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00211; color=pink; size=large; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "orange, medium", "candidate_1": "pink, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00211\",\"pink\",\"large\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 1, "candidate_0": 0}, "ready": {"false": 0.9299999999999999, "true": 0.07}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:99", "state_id": "catalog_lookup_v2:train:99", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00166; color=black; size=large; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "blue, large", "candidate_1": "brown, small", "candidate_2": "gray, large", "candidate_3": "purple, medium", "candidate_4": "blue, small", "candidate_5": "orange, small", "candidate_6": "red, large", "candidate_7": "cyan, medium", "candidate_8": "pink, small", "candidate_9": "black, large", "candidate_10": "brown, medium", "candidate_11": "green, large", "candidate_12": "orange, medium", "candidate_13": "blue, medium", "candidate_14": "yellow, small", "candidate_15": "pink, large", "candidate_16": "orange, large", "candidate_17": "gray, small", "candidate_18": "brown, large", "candidate_19": "purple, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_9", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 1, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00166\",\"black\",\"large\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_11": 0, "candidate_2": 0, "candidate_4": 0, "candidate_5": 0, "candidate_9": 1, "candidate_17": 0, "candidate_3": 0, "candidate_10": 0, "candidate_14": 0, "candidate_19": 0, "candidate_13": 0, "candidate_0": 0, "candidate_6": 0, "candidate_18": 0, "candidate_15": 0, "candidate_12": 0, "candidate_1": 0, "candidate_8": 0, "candidate_16": 0, "candidate_7": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:98", "state_id": "catalog_lookup_v2:train:98", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00209; color=pink; size=large; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "orange, medium", "candidate_1": "gray, large", "candidate_2": "gray, medium", "candidate_3": "white, medium", "candidate_4": "red, medium", "candidate_5": "yellow, large", "candidate_6": "black, large", "candidate_7": "pink, large", "candidate_8": "green, large", "candidate_9": "cyan, medium", "candidate_10": "purple, small", "candidate_11": "orange, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_7", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 1, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00209\",\"pink\",\"large\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_10": 0, "candidate_2": 0, "candidate_11": 0, "candidate_4": 0, "candidate_5": 0, "candidate_9": 0, "candidate_8": 0, "candidate_0": 0, "candidate_6": 0, "candidate_1": 0, "candidate_3": 0, "candidate_7": 1}, "ready": {"false": 0.8, "true": 0.2}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:97", "state_id": "catalog_lookup_v2:train:97", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00189; color=orange; size=large; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "pink, large", "candidate_1": "gray, medium", "candidate_2": "white, medium", "candidate_3": "orange, large", "candidate_4": "orange, small", "candidate_5": "cyan, small", "candidate_6": "green, small", "candidate_7": "brown, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_3", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 1, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00189\",\"orange\",\"large\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_5": 0, "candidate_4": 0, "candidate_7": 0, "candidate_1": 0, "candidate_0": 0, "candidate_3": 1, "candidate_6": 0, "candidate_2": 0}, "ready": {"false": 0.13, "true": 0.87}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:96", "state_id": "catalog_lookup_v2:train:96", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00168; color=orange; size=small; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "brown, small", "candidate_1": "red, small", "candidate_2": "white, large", "candidate_3": "orange, small", "candidate_4": "purple, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_3", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 1, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00168\",\"orange\",\"small\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_4": 0, "candidate_3": 1, "candidate_1": 0, "candidate_2": 0, "candidate_0": 0}, "ready": {"false": 0.77, "true": 0.23}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:100", "state_id": "catalog_lookup_v2:train:100", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00023; color=red; size=large; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "red, large", "candidate_1": "yellow, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00023\",\"red\",\"large\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:103", "state_id": "catalog_lookup_v2:train:103", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00179; color=orange; size=medium; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "red, small", "candidate_1": "orange, medium", "candidate_2": "brown, small", "candidate_3": "blue, small", "candidate_4": "cyan, small", "candidate_5": "red, large", "candidate_6": "brown, large", "candidate_7": "pink, medium", "candidate_8": "white, large", "candidate_9": "yellow, large", "candidate_10": "orange, small", "candidate_11": "white, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00179\",\"orange\",\"medium\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_4": 0, "candidate_1": 1, "candidate_6": 0, "candidate_9": 0, "candidate_0": 0, "candidate_2": 0, "candidate_8": 0, "candidate_7": 0, "candidate_10": 0, "candidate_11": 0, "candidate_3": 0, "candidate_5": 0}, "ready": {"false": 0.94, "true": 0.06}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:101", "state_id": "catalog_lookup_v2:train:101", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00048; color=green; size=small; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "black, large", "candidate_1": "white, medium", "candidate_2": "gray, medium", "candidate_3": "blue, medium", "candidate_4": "green, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_4", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00048\",\"green\",\"small\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_3": 0, "candidate_2": 0, "candidate_4": 1, "candidate_1": 0, "candidate_0": 0}, "ready": {"false": 0.77, "true": 0.23}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:102", "state_id": "catalog_lookup_v2:train:102", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00065; color=green; size=large; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "cyan, small", "candidate_1": "red, large", "candidate_2": "brown, small", "candidate_3": "gray, medium", "candidate_4": "yellow, medium", "candidate_5": "green, large", "candidate_6": "yellow, small", "candidate_7": "white, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_5", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 1, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00065\",\"green\",\"large\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_7": 0, "candidate_5": 1, "candidate_3": 0, "candidate_4": 0, "candidate_6": 0, "candidate_2": 0, "candidate_0": 0, "candidate_1": 0}, "ready": {"false": 0.8, "true": 0.2}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:104", "state_id": "catalog_lookup_v2:train:104", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00251; color=gray; size=medium; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "brown, large", "candidate_1": "red, large", "candidate_2": "gray, small", "candidate_3": "black, medium", "candidate_4": "orange, medium", "candidate_5": "gray, medium", "candidate_6": "purple, medium", "candidate_7": "yellow, small", "candidate_8": "blue, large", "candidate_9": "black, large", "candidate_10": "orange, large", "candidate_11": "green, medium", "candidate_12": "green, large", "candidate_13": "purple, large", "candidate_14": "cyan, small", "candidate_15": "cyan, medium", "candidate_16": "pink, large", "candidate_17": "black, small", "candidate_18": "pink, medium", "candidate_19": "orange, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_5", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 1, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00251\",\"gray\",\"medium\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_12": 0, "candidate_0": 0, "candidate_13": 0, "candidate_14": 0, "candidate_16": 0, "candidate_8": 0, "candidate_3": 0, "candidate_5": 1, "candidate_4": 0, "candidate_19": 0, "candidate_7": 0, "candidate_11": 0, "candidate_17": 0, "candidate_9": 0, "candidate_10": 0, "candidate_6": 0, "candidate_15": 0, "candidate_2": 0, "candidate_1": 0, "candidate_18": 0}, "ready": {"false": 0.9299999999999999, "true": 0.07}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:105", "state_id": "catalog_lookup_v2:train:105", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00273; color=cyan; size=medium; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "cyan, medium", "candidate_1": "gray, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00273\",\"cyan\",\"medium\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 0.81, "true": 0.19}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:106", "state_id": "catalog_lookup_v2:train:106", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00020; color=red; size=large; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "orange, large", "candidate_1": "blue, medium", "candidate_2": "pink, small", "candidate_3": "pink, medium", "candidate_4": "red, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_4", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 1}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00020\",\"red\",\"large\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_4": 1, "candidate_3": 0, "candidate_0": 0, "candidate_2": 0, "candidate_1": 0}, "ready": {"false": 0.16000000000000003, "true": 0.84}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:110", "state_id": "catalog_lookup_v2:train:110", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00113; color=purple; size=large; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, large", "candidate_1": "gray, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00113\",\"purple\",\"large\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_0": 1}, "ready": {"false": 0.8, "true": 0.2}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:107", "state_id": "catalog_lookup_v2:train:107", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00095; color=yellow; size=large; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "blue, medium", "candidate_1": "yellow, large", "candidate_2": "yellow, small", "candidate_3": "pink, medium", "candidate_4": "gray, small", "candidate_5": "purple, large", "candidate_6": "black, large", "candidate_7": "green, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00095\",\"yellow\",\"large\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_3": 0, "candidate_2": 0, "candidate_7": 0, "candidate_1": 1, "candidate_6": 0, "candidate_4": 0, "candidate_0": 0, "candidate_5": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:109", "state_id": "catalog_lookup_v2:train:109", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00157; color=black; size=medium; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, small", "candidate_1": "white, large", "candidate_2": "blue, small", "candidate_3": "blue, medium", "candidate_4": "white, small", "candidate_5": "black, small", "candidate_6": "yellow, medium", "candidate_7": "pink, medium", "candidate_8": "orange, medium", "candidate_9": "brown, small", "candidate_10": "orange, small", "candidate_11": "brown, large", "candidate_12": "gray, small", "candidate_13": "red, medium", "candidate_14": "black, medium", "candidate_15": "blue, large", "candidate_16": "green, small", "candidate_17": "red, large", "candidate_18": "purple, large", "candidate_19": "green, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_14", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 1, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00157\",\"black\",\"medium\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_15": 0, "candidate_13": 0, "candidate_2": 0, "candidate_8": 0, "candidate_0": 0, "candidate_5": 0, "candidate_11": 0, "candidate_4": 0, "candidate_1": 0, "candidate_10": 0, "candidate_18": 0, "candidate_3": 0, "candidate_16": 0, "candidate_7": 0, "candidate_17": 0, "candidate_19": 0, "candidate_14": 1, "candidate_6": 0, "candidate_9": 0, "candidate_12": 0}, "ready": {"false": 0.15000000000000002, "true": 0.85}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:113", "state_id": "catalog_lookup_v2:train:113", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00053; color=green; size=small; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, large", "candidate_1": "white, small", "candidate_2": "red, large", "candidate_3": "brown, medium", "candidate_4": "pink, medium", "candidate_5": "yellow, medium", "candidate_6": "green, medium", "candidate_7": "green, small", "candidate_8": "brown, small", "candidate_9": "blue, medium", "candidate_10": "green, large", "candidate_11": "brown, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_7", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 1, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00053\",\"green\",\"small\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_11": 0, "candidate_0": 0, "candidate_2": 0, "candidate_5": 0, "candidate_8": 0, "candidate_6": 0, "candidate_4": 0, "candidate_10": 0, "candidate_7": 1, "candidate_3": 0, "candidate_9": 0}, "ready": {"false": 0.14, "true": 0.86}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:112", "state_id": "catalog_lookup_v2:train:112", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00287; color=cyan; size=large; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "yellow, small", "candidate_1": "black, small", "candidate_2": "cyan, large", "candidate_3": "cyan, medium", "candidate_4": "blue, medium", "candidate_5": "brown, small", "candidate_6": "gray, small", "candidate_7": "green, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_2", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 1, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00287\",\"cyan\",\"large\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_5": 0, "candidate_4": 0, "candidate_7": 0, "candidate_1": 0, "candidate_0": 0, "candidate_3": 0, "candidate_6": 0, "candidate_2": 1}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:111", "state_id": "catalog_lookup_v2:train:111", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00207; color=pink; size=medium; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "blue, large", "candidate_1": "green, large", "candidate_2": "green, small", "candidate_3": "white, medium", "candidate_4": "pink, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_4", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00207\",\"pink\",\"medium\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_4": 1, "candidate_1": 0, "candidate_0": 0, "candidate_3": 0, "candidate_2": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:114", "state_id": "catalog_lookup_v2:train:114", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00064; color=green; size=large; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "green, large", "candidate_1": "black, small", "candidate_2": "black, medium", "candidate_3": "gray, medium", "candidate_4": "purple, large", "candidate_5": "blue, medium", "candidate_6": "white, small", "candidate_7": "orange, small", "candidate_8": "yellow, medium", "candidate_9": "brown, small", "candidate_10": "yellow, small", "candidate_11": "orange, medium", "candidate_12": "cyan, medium", "candidate_13": "black, large", "candidate_14": "gray, small", "candidate_15": "orange, large", "candidate_16": "white, large", "candidate_17": "brown, medium", "candidate_18": "pink, small", "candidate_19": "purple, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00064\",\"green\",\"large\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_18": 0, "candidate_16": 0, "candidate_2": 0, "candidate_0": 1, "candidate_19": 0, "candidate_1": 0, "candidate_8": 0, "candidate_13": 0, "candidate_6": 0, "candidate_7": 0, "candidate_15": 0, "candidate_9": 0, "candidate_3": 0, "candidate_14": 0, "candidate_4": 0, "candidate_10": 0, "candidate_11": 0, "candidate_17": 0, "candidate_5": 0, "candidate_12": 0}, "ready": {"false": 0.76, "true": 0.24}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:108", "state_id": "catalog_lookup_v2:train:108", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00038; color=blue; size=medium; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "blue, small", "candidate_1": "green, small", "candidate_2": "cyan, small", "candidate_3": "purple, small", "candidate_4": "brown, small", "candidate_5": "gray, medium", "candidate_6": "orange, medium", "candidate_7": "blue, medium", "candidate_8": "green, large", "candidate_9": "orange, large", "candidate_10": "black, small", "candidate_11": "brown, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_7", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 1, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00038\",\"blue\",\"medium\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_4": 0, "candidate_11": 0, "candidate_10": 0, "candidate_6": 0, "candidate_3": 0, "candidate_5": 0, "candidate_8": 0, "candidate_9": 0, "candidate_0": 0, "candidate_2": 0, "candidate_7": 1}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:115", "state_id": "catalog_lookup_v2:train:115", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00005; color=red; size=small; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "red, small", "candidate_1": "purple, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00005\",\"red\",\"small\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_0": 1}, "ready": {"false": 0.13, "true": 0.87}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:117", "state_id": "catalog_lookup_v2:train:117", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00266; color=cyan; size=small; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "green, medium", "candidate_1": "orange, medium", "candidate_2": "green, small", "candidate_3": "cyan, medium", "candidate_4": "yellow, medium", "candidate_5": "blue, large", "candidate_6": "cyan, large", "candidate_7": "cyan, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_7", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00266\",\"cyan\",\"small\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 0, "candidate_6": 0, "candidate_1": 0, "candidate_2": 0, "candidate_5": 0, "candidate_3": 0, "candidate_4": 0, "candidate_7": 1}, "ready": {"false": 0.92, "true": 0.08}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:116", "state_id": "catalog_lookup_v2:train:116", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00075; color=yellow; size=small; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "yellow, small", "candidate_1": "green, medium", "candidate_2": "red, large", "candidate_3": "green, large", "candidate_4": "pink, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00075\",\"yellow\",\"small\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 1, "candidate_2": 0, "candidate_4": 0, "candidate_1": 0, "candidate_3": 0}, "ready": {"false": 0.92, "true": 0.08}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:120", "state_id": "catalog_lookup_v2:train:120", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00228; color=brown; size=medium; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "brown, medium", "candidate_1": "white, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00228\",\"brown\",\"medium\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 0.16000000000000003, "true": 0.84}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:119", "state_id": "catalog_lookup_v2:train:119", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00255; color=gray; size=medium; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "cyan, large", "candidate_1": "green, small", "candidate_2": "black, small", "candidate_3": "yellow, large", "candidate_4": "yellow, medium", "candidate_5": "purple, small", "candidate_6": "yellow, small", "candidate_7": "green, large", "candidate_8": "pink, large", "candidate_9": "gray, large", "candidate_10": "pink, small", "candidate_11": "cyan, small", "candidate_12": "brown, small", "candidate_13": "pink, medium", "candidate_14": "red, medium", "candidate_15": "blue, large", "candidate_16": "red, small", "candidate_17": "gray, medium", "candidate_18": "black, medium", "candidate_19": "brown, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_17", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 1, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00255\",\"gray\",\"medium\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_18": 0, "candidate_5": 0, "candidate_12": 0, "candidate_4": 0, "candidate_7": 0, "candidate_15": 0, "candidate_3": 0, "candidate_6": 0, "candidate_14": 0, "candidate_13": 0, "candidate_2": 0, "candidate_11": 0, "candidate_17": 1, "candidate_1": 0, "candidate_16": 0, "candidate_8": 0, "candidate_10": 0, "candidate_0": 0, "candidate_19": 0, "candidate_9": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:121", "state_id": "catalog_lookup_v2:train:121", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00160; color=black; size=large; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "red, large", "candidate_1": "orange, small", "candidate_2": "black, large", "candidate_3": "gray, small", "candidate_4": "pink, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_2", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 1, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00160\",\"black\",\"large\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_3": 0, "candidate_2": 1, "candidate_0": 0, "candidate_4": 0}, "ready": {"false": 0.78, "true": 0.22}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:122", "state_id": "catalog_lookup_v2:train:122", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00182; color=orange; size=medium; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "orange, medium", "candidate_1": "cyan, medium", "candidate_2": "blue, large", "candidate_3": "red, medium", "candidate_4": "gray, medium", "candidate_5": "yellow, large", "candidate_6": "green, small", "candidate_7": "cyan, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00182\",\"orange\",\"medium\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_3": 0, "candidate_0": 1, "candidate_6": 0, "candidate_4": 0, "candidate_5": 0, "candidate_1": 0, "candidate_2": 0, "candidate_7": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:124", "state_id": "catalog_lookup_v2:train:124", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00236; color=brown; size=large; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "yellow, large", "candidate_1": "brown, small", "candidate_2": "purple, small", "candidate_3": "green, large", "candidate_4": "blue, medium", "candidate_5": "red, medium", "candidate_6": "gray, small", "candidate_7": "blue, small", "candidate_8": "orange, small", "candidate_9": "black, medium", "candidate_10": "pink, large", "candidate_11": "white, small", "candidate_12": "brown, medium", "candidate_13": "gray, medium", "candidate_14": "white, medium", "candidate_15": "blue, large", "candidate_16": "white, large", "candidate_17": "brown, large", "candidate_18": "purple, large", "candidate_19": "gray, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_17", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 1, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00236\",\"brown\",\"large\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 0, "candidate_3": 0, "candidate_18": 0, "candidate_4": 0, "candidate_13": 0, "candidate_6": 0, "candidate_17": 1, "candidate_11": 0, "candidate_19": 0, "candidate_16": 0, "candidate_14": 0, "candidate_8": 0, "candidate_9": 0, "candidate_5": 0, "candidate_7": 0, "candidate_10": 0, "candidate_15": 0, "candidate_2": 0, "candidate_12": 0, "candidate_1": 0}, "ready": {"false": 0.18000000000000005, "true": 0.82}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:123", "state_id": "catalog_lookup_v2:train:123", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00056; color=green; size=medium; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "cyan, large", "candidate_1": "pink, large", "candidate_2": "purple, large", "candidate_3": "pink, small", "candidate_4": "white, large", "candidate_5": "yellow, small", "candidate_6": "orange, large", "candidate_7": "blue, large", "candidate_8": "green, medium", "candidate_9": "brown, small", "candidate_10": "blue, small", "candidate_11": "black, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_8", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 1, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00056\",\"green\",\"medium\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_10": 0, "candidate_11": 0, "candidate_6": 0, "candidate_2": 0, "candidate_8": 1, "candidate_5": 0, "candidate_4": 0, "candidate_0": 0, "candidate_7": 0, "candidate_3": 0, "candidate_9": 0}, "ready": {"false": 0.77, "true": 0.23}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:125", "state_id": "catalog_lookup_v2:train:125", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00190; color=orange; size=large; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "orange, large", "candidate_1": "blue, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00190\",\"orange\",\"large\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_0": 1}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:118", "state_id": "catalog_lookup_v2:train:118", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00220; color=brown; size=small; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, small", "candidate_1": "purple, medium", "candidate_2": "pink, medium", "candidate_3": "yellow, large", "candidate_4": "yellow, medium", "candidate_5": "brown, small", "candidate_6": "red, medium", "candidate_7": "brown, medium", "candidate_8": "pink, small", "candidate_9": "white, large", "candidate_10": "black, large", "candidate_11": "blue, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_5", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 1, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00220\",\"brown\",\"small\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_8": 0, "candidate_3": 0, "candidate_4": 0, "candidate_9": 0, "candidate_10": 0, "candidate_7": 0, "candidate_1": 0, "candidate_5": 1, "candidate_6": 0, "candidate_0": 0, "candidate_11": 0, "candidate_2": 0}, "ready": {"false": 0.17000000000000004, "true": 0.83}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:126", "state_id": "catalog_lookup_v2:train:126", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00011; color=red; size=medium; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "green, large", "candidate_1": "red, medium", "candidate_2": "pink, medium", "candidate_3": "white, small", "candidate_4": "yellow, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00011\",\"red\",\"medium\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_4": 0, "candidate_3": 0, "candidate_0": 0, "candidate_2": 0, "candidate_1": 1}, "ready": {"false": 0.94, "true": 0.06}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:127", "state_id": "catalog_lookup_v2:train:127", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00197; color=pink; size=small; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "brown, medium", "candidate_1": "red, small", "candidate_2": "black, large", "candidate_3": "pink, small", "candidate_4": "yellow, large", "candidate_5": "white, small", "candidate_6": "cyan, large", "candidate_7": "green, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_3", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 1, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00197\",\"pink\",\"small\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_4": 0, "candidate_1": 0, "candidate_6": 0, "candidate_0": 0, "candidate_2": 0, "candidate_7": 0, "candidate_3": 1, "candidate_5": 0}, "ready": {"false": 0.13, "true": 0.87}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:128", "state_id": "catalog_lookup_v2:train:128", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00003; color=red; size=small; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, small", "candidate_1": "purple, large", "candidate_2": "red, large", "candidate_3": "blue, medium", "candidate_4": "gray, large", "candidate_5": "brown, small", "candidate_6": "white, small", "candidate_7": "gray, medium", "candidate_8": "red, small", "candidate_9": "black, small", "candidate_10": "green, medium", "candidate_11": "green, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_8", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 1, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00003\",\"red\",\"small\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 0, "candidate_10": 0, "candidate_6": 0, "candidate_1": 0, "candidate_9": 0, "candidate_2": 0, "candidate_8": 1, "candidate_5": 0, "candidate_3": 0, "candidate_7": 0, "candidate_4": 0, "candidate_11": 0}, "ready": {"false": 0.9299999999999999, "true": 0.07}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:129", "state_id": "catalog_lookup_v2:train:129", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00237; color=brown; size=large; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "gray, small", "candidate_1": "yellow, large", "candidate_2": "gray, medium", "candidate_3": "green, large", "candidate_4": "white, large", "candidate_5": "brown, small", "candidate_6": "cyan, small", "candidate_7": "white, medium", "candidate_8": "red, medium", "candidate_9": "purple, large", "candidate_10": "gray, large", "candidate_11": "green, medium", "candidate_12": "brown, medium", "candidate_13": "cyan, large", "candidate_14": "pink, large", "candidate_15": "cyan, medium", "candidate_16": "orange, medium", "candidate_17": "brown, large", "candidate_18": "green, small", "candidate_19": "red, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_17", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 1, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00237\",\"brown\",\"large\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_9": 0, "candidate_5": 0, "candidate_0": 0, "candidate_14": 0, "candidate_8": 0, "candidate_1": 0, "candidate_12": 0, "candidate_11": 0, "candidate_18": 0, "candidate_10": 0, "candidate_15": 0, "candidate_2": 0, "candidate_16": 0, "candidate_17": 1, "candidate_19": 0, "candidate_3": 0, "candidate_4": 0, "candidate_13": 0, "candidate_7": 0, "candidate_6": 0}, "ready": {"false": 0.13, "true": 0.87}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:130", "state_id": "catalog_lookup_v2:train:130", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00032; color=blue; size=medium; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "blue, medium", "candidate_1": "black, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00032\",\"blue\",\"medium\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 0.78, "true": 0.22}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:133", "state_id": "catalog_lookup_v2:train:133", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00274; color=cyan; size=medium; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, medium", "candidate_1": "red, large", "candidate_2": "red, medium", "candidate_3": "cyan, medium", "candidate_4": "cyan, large", "candidate_5": "blue, large", "candidate_6": "black, small", "candidate_7": "gray, large", "candidate_8": "white, small", "candidate_9": "green, medium", "candidate_10": "pink, small", "candidate_11": "blue, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_3", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 1, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00274\",\"cyan\",\"medium\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_3": 1, "candidate_2": 0, "candidate_11": 0, "candidate_4": 0, "candidate_5": 0, "candidate_9": 0, "candidate_6": 0, "candidate_0": 0, "candidate_8": 0, "candidate_1": 0, "candidate_10": 0, "candidate_7": 0}, "ready": {"false": 0.92, "true": 0.08}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:135", "state_id": "catalog_lookup_v2:train:135", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00214; color=pink; size=large; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "blue, medium", "candidate_1": "pink, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00214\",\"pink\",\"large\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:137", "state_id": "catalog_lookup_v2:train:137", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00249; color=gray; size=medium; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, large", "candidate_1": "blue, small", "candidate_2": "red, medium", "candidate_3": "purple, medium", "candidate_4": "brown, small", "candidate_5": "pink, small", "candidate_6": "yellow, small", "candidate_7": "gray, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_7", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00249\",\"gray\",\"medium\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_6": 0, "candidate_2": 0, "candidate_5": 0, "candidate_4": 0, "candidate_0": 0, "candidate_7": 1, "candidate_3": 0}, "ready": {"false": 0.79, "true": 0.21}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:134", "state_id": "catalog_lookup_v2:train:134", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00100; color=purple; size=small; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "green, medium", "candidate_1": "black, small", "candidate_2": "cyan, small", "candidate_3": "yellow, large", "candidate_4": "white, large", "candidate_5": "purple, small", "candidate_6": "blue, small", "candidate_7": "cyan, large", "candidate_8": "pink, small", "candidate_9": "green, large", "candidate_10": "brown, medium", "candidate_11": "cyan, medium", "candidate_12": "white, medium", "candidate_13": "pink, medium", "candidate_14": "brown, large", "candidate_15": "gray, medium", "candidate_16": "white, small", "candidate_17": "pink, large", "candidate_18": "orange, medium", "candidate_19": "purple, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_5", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 1, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00100\",\"purple\",\"small\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_2": 0, "candidate_0": 0, "candidate_7": 0, "candidate_18": 0, "candidate_12": 0, "candidate_11": 0, "candidate_13": 0, "candidate_10": 0, "candidate_5": 1, "candidate_9": 0, "candidate_19": 0, "candidate_1": 0, "candidate_4": 0, "candidate_6": 0, "candidate_17": 0, "candidate_8": 0, "candidate_3": 0}, "ready": {"false": 0.16000000000000003, "true": 0.84}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:138", "state_id": "catalog_lookup_v2:train:138", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00097; color=purple; size=small; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "orange, large", "candidate_1": "purple, small", "candidate_2": "white, small", "candidate_3": "pink, medium", "candidate_4": "orange, medium", "candidate_5": "blue, medium", "candidate_6": "purple, medium", "candidate_7": "pink, large", "candidate_8": "red, small", "candidate_9": "purple, large", "candidate_10": "orange, small", "candidate_11": "pink, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00097\",\"purple\",\"small\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 1, "candidate_0": 0, "candidate_3": 0, "candidate_6": 0, "candidate_2": 0, "candidate_8": 0, "candidate_5": 0, "candidate_4": 0, "candidate_10": 0, "candidate_7": 0, "candidate_11": 0, "candidate_9": 0}, "ready": {"false": 0.8, "true": 0.2}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:136", "state_id": "catalog_lookup_v2:train:136", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00150; color=black; size=small; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "gray, large", "candidate_1": "purple, medium", "candidate_2": "black, small", "candidate_3": "orange, small", "candidate_4": "white, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_2", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 1, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00150\",\"black\",\"small\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 0, "candidate_2": 1, "candidate_4": 0, "candidate_1": 0, "candidate_3": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:132", "state_id": "catalog_lookup_v2:train:132", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00186; color=orange; size=large; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "pink, large", "candidate_1": "white, small", "candidate_2": "green, medium", "candidate_3": "yellow, medium", "candidate_4": "cyan, large", "candidate_5": "brown, medium", "candidate_6": "orange, large", "candidate_7": "red, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_6", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 1, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00186\",\"orange\",\"large\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_5": 0, "candidate_1": 0, "candidate_7": 0, "candidate_4": 0, "candidate_0": 0, "candidate_3": 0, "candidate_6": 1, "candidate_2": 0}, "ready": {"false": 0.9299999999999999, "true": 0.07}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:139", "state_id": "catalog_lookup_v2:train:139", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00206; color=pink; size=medium; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "red, large", "candidate_1": "green, medium", "candidate_2": "orange, large", "candidate_3": "green, large", "candidate_4": "brown, large", "candidate_5": "brown, small", "candidate_6": "black, medium", "candidate_7": "cyan, medium", "candidate_8": "blue, medium", "candidate_9": "white, large", "candidate_10": "green, small", "candidate_11": "blue, large", "candidate_12": "cyan, small", "candidate_13": "pink, medium", "candidate_14": "white, medium", "candidate_15": "yellow, medium", "candidate_16": "black, large", "candidate_17": "purple, small", "candidate_18": "pink, small", "candidate_19": "red, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_13", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 1, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00206\",\"pink\",\"medium\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_19": 0, "candidate_13": 1, "candidate_12": 0, "candidate_8": 0, "candidate_17": 0, "candidate_16": 0, "candidate_4": 0, "candidate_14": 0, "candidate_10": 0, "candidate_1": 0, "candidate_11": 0, "candidate_3": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_0": 0, "candidate_15": 0, "candidate_2": 0, "candidate_18": 0, "candidate_9": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:143", "state_id": "catalog_lookup_v2:train:143", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00253; color=gray; size=medium; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "blue, small", "candidate_1": "gray, large", "candidate_2": "yellow, medium", "candidate_3": "black, large", "candidate_4": "white, small", "candidate_5": "red, medium", "candidate_6": "gray, small", "candidate_7": "cyan, medium", "candidate_8": "pink, small", "candidate_9": "black, small", "candidate_10": "gray, medium", "candidate_11": "brown, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_10", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 1, "candidate_11": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00253\",\"gray\",\"medium\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_3": 0, "candidate_6": 0, "candidate_7": 0, "candidate_2": 0, "candidate_8": 0, "candidate_0": 0, "candidate_5": 0, "candidate_11": 0, "candidate_9": 0, "candidate_1": 0, "candidate_4": 0, "candidate_10": 1}, "ready": {"false": 0.14, "true": 0.86}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:141", "state_id": "catalog_lookup_v2:train:141", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00217; color=brown; size=small; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "gray, medium", "candidate_1": "purple, small", "candidate_2": "brown, small", "candidate_3": "cyan, small", "candidate_4": "yellow, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_2", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 1, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00217\",\"brown\",\"small\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_3": 0, "candidate_4": 0, "candidate_0": 0, "candidate_2": 1}, "ready": {"false": 0.78, "true": 0.22}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:140", "state_id": "catalog_lookup_v2:train:140", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00143; color=white; size=large; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "white, large", "candidate_1": "orange, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00143\",\"white\",\"large\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:144", "state_id": "catalog_lookup_v2:train:144", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00219; color=brown; size=small; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, medium", "candidate_1": "white, medium", "candidate_2": "black, medium", "candidate_3": "white, small", "candidate_4": "green, small", "candidate_5": "gray, large", "candidate_6": "yellow, medium", "candidate_7": "black, large", "candidate_8": "black, small", "candidate_9": "red, large", "candidate_10": "brown, medium", "candidate_11": "purple, large", "candidate_12": "green, large", "candidate_13": "white, large", "candidate_14": "yellow, large", "candidate_15": "brown, small", "candidate_16": "orange, large", "candidate_17": "pink, small", "candidate_18": "yellow, small", "candidate_19": "blue, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_15", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 1, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00219\",\"brown\",\"small\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_14": 0, "candidate_18": 0, "candidate_8": 0, "candidate_13": 0, "candidate_7": 0, "candidate_1": 0, "candidate_19": 0, "candidate_12": 0, "candidate_0": 0, "candidate_6": 0, "candidate_2": 0, "candidate_11": 0, "candidate_16": 0, "candidate_4": 0, "candidate_3": 0, "candidate_9": 0, "candidate_10": 0, "candidate_5": 0, "candidate_15": 1, "candidate_17": 0}, "ready": {"false": 0.9299999999999999, "true": 0.07}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:142", "state_id": "catalog_lookup_v2:train:142", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00024; color=blue; size=small; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "gray, medium", "candidate_1": "orange, large", "candidate_2": "black, medium", "candidate_3": "purple, medium", "candidate_4": "purple, small", "candidate_5": "red, medium", "candidate_6": "cyan, small", "candidate_7": "blue, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_7", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00024\",\"blue\",\"small\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_2": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_0": 0, "candidate_1": 0, "candidate_3": 0, "candidate_7": 1}, "ready": {"false": 0.77, "true": 0.23}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:146", "state_id": "catalog_lookup_v2:train:146", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00131; color=white; size=medium; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "pink, large", "candidate_1": "white, medium", "candidate_2": "purple, medium", "candidate_3": "yellow, large", "candidate_4": "cyan, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00131\",\"white\",\"medium\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 1, "candidate_3": 0, "candidate_2": 0, "candidate_0": 0, "candidate_4": 0}, "ready": {"false": 0.91, "true": 0.09}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:131", "state_id": "catalog_lookup_v2:train:131", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00060; color=green; size=medium; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "white, medium", "candidate_1": "green, medium", "candidate_2": "gray, medium", "candidate_3": "orange, medium", "candidate_4": "brown, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00060\",\"green\",\"medium\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_4": 0, "candidate_2": 0, "candidate_1": 1, "candidate_3": 0, "candidate_0": 0}, "ready": {"false": 0.16000000000000003, "true": 0.84}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:145", "state_id": "catalog_lookup_v2:train:145", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00171; color=orange; size=small; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "black, medium", "candidate_1": "orange, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00171\",\"orange\",\"small\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 0.9299999999999999, "true": 0.07}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:147", "state_id": "catalog_lookup_v2:train:147", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00134; color=white; size=medium; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, small", "candidate_1": "green, large", "candidate_2": "pink, medium", "candidate_3": "pink, small", "candidate_4": "black, medium", "candidate_5": "black, small", "candidate_6": "white, medium", "candidate_7": "cyan, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_6", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 1, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00134\",\"white\",\"medium\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_4": 0, "candidate_2": 0, "candidate_3": 0, "candidate_5": 0, "candidate_6": 1, "candidate_1": 0, "candidate_7": 0, "candidate_0": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:149", "state_id": "catalog_lookup_v2:train:149", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00101; color=purple; size=small; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "gray, medium", "candidate_1": "green, medium", "candidate_2": "blue, large", "candidate_3": "cyan, medium", "candidate_4": "purple, medium", "candidate_5": "pink, large", "candidate_6": "green, small", "candidate_7": "cyan, large", "candidate_8": "orange, small", "candidate_9": "blue, medium", "candidate_10": "yellow, small", "candidate_11": "red, small", "candidate_12": "brown, small", "candidate_13": "white, large", "candidate_14": "gray, large", "candidate_15": "brown, medium", "candidate_16": "purple, small", "candidate_17": "black, small", "candidate_18": "green, large", "candidate_19": "white, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_16", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 1, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00101\",\"purple\",\"small\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_18": 0, "candidate_9": 0, "candidate_16": 1, "candidate_12": 0, "candidate_4": 0, "candidate_19": 0, "candidate_1": 0, "candidate_14": 0, "candidate_2": 0, "candidate_15": 0, "candidate_11": 0, "candidate_8": 0, "candidate_7": 0, "candidate_17": 0, "candidate_3": 0, "candidate_0": 0, "candidate_6": 0, "candidate_5": 0, "candidate_10": 0, "candidate_13": 0}, "ready": {"false": 0.12, "true": 0.88}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:148", "state_id": "catalog_lookup_v2:train:148", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00238; color=brown; size=large; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "yellow, large", "candidate_1": "purple, medium", "candidate_2": "brown, small", "candidate_3": "yellow, small", "candidate_4": "white, large", "candidate_5": "gray, medium", "candidate_6": "cyan, medium", "candidate_7": "green, small", "candidate_8": "green, large", "candidate_9": "red, small", "candidate_10": "black, large", "candidate_11": "brown, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_11", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00238\",\"brown\",\"large\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_8": 0, "candidate_4": 0, "candidate_9": 0, "candidate_5": 0, "candidate_10": 0, "candidate_2": 0, "candidate_3": 0, "candidate_7": 0, "candidate_11": 1, "candidate_0": 0, "candidate_1": 0, "candidate_6": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:150", "state_id": "catalog_lookup_v2:train:150", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00034; color=blue; size=medium; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "white, large", "candidate_1": "blue, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00034\",\"blue\",\"medium\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 1, "candidate_0": 0}, "ready": {"false": 0.9299999999999999, "true": 0.07}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:152", "state_id": "catalog_lookup_v2:train:152", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00202; color=pink; size=medium; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "yellow, large", "candidate_1": "green, medium", "candidate_2": "white, small", "candidate_3": "green, small", "candidate_4": "blue, large", "candidate_5": "pink, medium", "candidate_6": "cyan, large", "candidate_7": "gray, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_5", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 1, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00202\",\"pink\",\"medium\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_5": 1, "candidate_4": 0, "candidate_7": 0, "candidate_1": 0, "candidate_0": 0, "candidate_3": 0, "candidate_6": 0, "candidate_2": 0}, "ready": {"false": 0.9299999999999999, "true": 0.07}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:151", "state_id": "catalog_lookup_v2:train:151", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00155; color=black; size=medium; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "pink, small", "candidate_1": "white, small", "candidate_2": "green, large", "candidate_3": "cyan, large", "candidate_4": "black, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_4", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00155\",\"black\",\"medium\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_4": 1, "candidate_2": 0, "candidate_3": 0, "candidate_0": 0, "candidate_1": 0}, "ready": {"false": 0.9299999999999999, "true": 0.07}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:153", "state_id": "catalog_lookup_v2:train:153", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00210; color=pink; size=large; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "yellow, small", "candidate_1": "purple, medium", "candidate_2": "pink, large", "candidate_3": "black, medium", "candidate_4": "brown, small", "candidate_5": "blue, large", "candidate_6": "yellow, large", "candidate_7": "black, small", "candidate_8": "white, medium", "candidate_9": "cyan, small", "candidate_10": "orange, medium", "candidate_11": "cyan, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_2", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 1, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00210\",\"pink\",\"large\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 0, "candidate_8": 0, "candidate_11": 0, "candidate_1": 0, "candidate_7": 0, "candidate_5": 0, "candidate_6": 0, "candidate_4": 0, "candidate_9": 0, "candidate_3": 0, "candidate_10": 0, "candidate_2": 1}, "ready": {"false": 0.92, "true": 0.08}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:154", "state_id": "catalog_lookup_v2:train:154", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00063; color=green; size=medium; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "cyan, small", "candidate_1": "black, large", "candidate_2": "blue, small", "candidate_3": "red, medium", "candidate_4": "green, medium", "candidate_5": "cyan, medium", "candidate_6": "purple, medium", "candidate_7": "brown, large", "candidate_8": "yellow, large", "candidate_9": "orange, medium", "candidate_10": "yellow, small", "candidate_11": "pink, small", "candidate_12": "pink, large", "candidate_13": "red, large", "candidate_14": "black, medium", "candidate_15": "brown, medium", "candidate_16": "green, small", "candidate_17": "red, small", "candidate_18": "blue, medium", "candidate_19": "brown, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_4", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 1, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00063\",\"green\",\"medium\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_18": 0, "candidate_5": 0, "candidate_12": 0, "candidate_4": 1, "candidate_7": 0, "candidate_15": 0, "candidate_3": 0, "candidate_6": 0, "candidate_14": 0, "candidate_13": 0, "candidate_2": 0, "candidate_11": 0, "candidate_17": 0, "candidate_1": 0, "candidate_16": 0, "candidate_8": 0, "candidate_10": 0, "candidate_0": 0, "candidate_19": 0, "candidate_9": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:155", "state_id": "catalog_lookup_v2:train:155", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00130; color=white; size=medium; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "brown, large", "candidate_1": "white, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00130\",\"white\",\"medium\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 1, "candidate_0": 0}, "ready": {"false": 0.92, "true": 0.08}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:156", "state_id": "catalog_lookup_v2:train:156", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00008; color=red; size=medium; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "green, medium", "candidate_1": "red, medium", "candidate_2": "yellow, small", "candidate_3": "black, large", "candidate_4": "green, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00008\",\"red\",\"medium\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_0": 0, "candidate_4": 0}, "ready": {"false": 0.77, "true": 0.23}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:157", "state_id": "catalog_lookup_v2:train:157", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00012; color=red; size=medium; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "white, medium", "candidate_1": "gray, small", "candidate_2": "black, small", "candidate_3": "green, small", "candidate_4": "purple, medium", "candidate_5": "purple, large", "candidate_6": "red, medium", "candidate_7": "green, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_6", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 1, "candidate_7": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00012\",\"red\",\"medium\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_5": 0, "candidate_2": 0, "candidate_6": 1, "candidate_4": 0, "candidate_0": 0, "candidate_7": 0, "candidate_3": 0}, "ready": {"false": 0.16000000000000003, "true": 0.84}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:158", "state_id": "catalog_lookup_v2:train:158", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00178; color=orange; size=medium; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "gray, medium", "candidate_1": "purple, large", "candidate_2": "orange, medium", "candidate_3": "black, small", "candidate_4": "purple, medium", "candidate_5": "blue, small", "candidate_6": "brown, medium", "candidate_7": "pink, medium", "candidate_8": "white, small", "candidate_9": "pink, large", "candidate_10": "gray, small", "candidate_11": "red, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_2", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 1, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00178\",\"orange\",\"medium\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_8": 0, "candidate_4": 0, "candidate_9": 0, "candidate_5": 0, "candidate_10": 0, "candidate_2": 1, "candidate_3": 0, "candidate_7": 0, "candidate_11": 0, "candidate_0": 0, "candidate_1": 0, "candidate_6": 0}, "ready": {"false": 0.9299999999999999, "true": 0.07}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:2", "state_id": "known_chance_v2:train:2", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: gray=2, white=9, red=5, black=9, pink=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"gray": "The sampled ball is gray", "pink": "The sampled ball is pink", "white": "The sampled ball is white", "black": "The sampled ball is black", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: gray, black", "The drawn color is one of: white, pink", "The drawn color is one of: red"]}}, "gold": {}, "gold_probs": {"outcome": {"gray": 0.0625, "pink": 0.21875, "white": 0.28125, "black": 0.28125, "red": 0.15625}, "focus": {"false": 0.71875, "true": 0.28125}, "group": {"0": 0.34375, "1": 0.5, "2": 0.15625}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"9/32\"],[\"gray\",\"1/16\"],[\"pink\",\"7/32\"],[\"red\",\"5/32\"],[\"white\",\"9/32\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0.8099999999999999, "black": 0.12, "pink": 0.01, "red": 0.01, "gray": 0.05}, "focus": {"false": 0.78, "true": 0.22}, "group": {"0": 0.21, "1": 0.78, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:train:159", "state_id": "catalog_lookup_v2:train:159", "family_id": "catalog_lookup_v2", "split": "train", "state": "Product record: item=SKU-00239; color=brown; size=large; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "red, small", "candidate_1": "white, large", "candidate_2": "red, large", "candidate_3": "purple, large", "candidate_4": "purple, small", "candidate_5": "blue, large", "candidate_6": "green, small", "candidate_7": "blue, small", "candidate_8": "blue, medium", "candidate_9": "cyan, small", "candidate_10": "black, large", "candidate_11": "brown, large", "candidate_12": "orange, large", "candidate_13": "white, medium", "candidate_14": "green, large", "candidate_15": "brown, medium", "candidate_16": "cyan, medium", "candidate_17": "gray, medium", "candidate_18": "red, medium", "candidate_19": "black, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_11", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 1, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00239\",\"brown\",\"large\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_13": 0, "candidate_0": 0, "candidate_7": 0, "candidate_18": 0, "candidate_12": 0, "candidate_11": 1, "candidate_10": 0, "candidate_9": 0, "candidate_5": 0, "candidate_2": 0, "candidate_4": 0, "candidate_1": 0, "candidate_19": 0, "candidate_6": 0, "candidate_17": 0, "candidate_8": 0, "candidate_3": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:0", "state_id": "known_chance_v2:train:0", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: green=7, purple=5, brown=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"brown": "The sampled ball is brown", "purple": "The sampled ball is purple", "green": "The sampled ball is green"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a purple ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: green", "The drawn color is one of: purple", "The drawn color is one of: brown"]}}, "gold": {}, "gold_probs": {"outcome": {"brown": 0.25, "purple": 0.3125, "green": 0.4375}, "focus": {"false": 0.6875, "true": 0.3125}, "group": {"0": 0.4375, "1": 0.3125, "2": 0.25}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"1/4\"],[\"green\",\"7/16\"],[\"purple\",\"5/16\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0.98, "brown": 0.01, "purple": 0.01}, "focus": {"false": 0.76, "true": 0.24}, "group": {"0": 0.96, "1": 0.02, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:1", "state_id": "known_chance_v2:train:1", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: white=2, red=7, black=7, brown=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"white": "The sampled ball is white", "black": "The sampled ball is black", "brown": "The sampled ball is brown", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a red ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white, brown", "The drawn color is one of: red", "The drawn color is one of: black"]}}, "gold": {}, "gold_probs": {"outcome": {"white": 0.08695652173913043, "black": 0.30434782608695654, "brown": 0.30434782608695654, "red": 0.30434782608695654}, "focus": {"false": 0.6956521739130435, "true": 0.30434782608695654}, "group": {"0": 0.391304347826087, "1": 0.30434782608695654, "2": 0.30434782608695654}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"7/23\"],[\"brown\",\"7/23\"],[\"red\",\"7/23\"],[\"white\",\"2/23\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"red": 0.45, "brown": 0.01, "white": 0.46, "black": 0.08}, "focus": {"false": 0.71, "true": 0.29}, "group": {"0": 0.16, "1": 0.64, "2": 0.2}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:3", "state_id": "known_chance_v2:train:3", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: gray=7, brown=2, blue=2, yellow=8, black=7, purple=4, pink=1, green=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "gray": "The sampled ball is gray", "yellow": "The sampled ball is yellow", "pink": "The sampled ball is pink", "green": "The sampled ball is green", "brown": "The sampled ball is brown", "purple": "The sampled ball is purple", "blue": "The sampled ball is blue"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a purple ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: gray, yellow, pink", "The drawn color is one of: brown, black, green", "The drawn color is one of: blue, purple"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.20588235294117646, "gray": 0.20588235294117646, "yellow": 0.23529411764705882, "pink": 0.029411764705882353, "green": 0.08823529411764706, "brown": 0.058823529411764705, "purple": 0.11764705882352941, "blue": 0.058823529411764705}, "focus": {"false": 0.8823529411764706, "true": 0.11764705882352941}, "group": {"0": 0.47058823529411764, "1": 0.3529411764705882, "2": 0.1764705882352941}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"7/34\"],[\"blue\",\"1/17\"],[\"brown\",\"1/17\"],[\"gray\",\"7/34\"],[\"green\",\"3/34\"],[\"pink\",\"1/34\"],[\"purple\",\"2/17\"],[\"yellow\",\"4/17\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.02, "green": 0, "gray": 0.06, "yellow": 0.91, "purple": 0, "blue": 0, "brown": 0, "black": 0.01}, "focus": {"false": 0.87, "true": 0.13}, "group": {"0": 0.99, "1": 0.01, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:4", "state_id": "known_chance_v2:train:4", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: brown=7, purple=4, cyan=3, red=1, green=7, blue=5, black=6, orange=6, white=9, pink=2, gray=3, yellow=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"red": "The sampled ball is red", "cyan": "The sampled ball is cyan", "blue": "The sampled ball is blue", "white": "The sampled ball is white", "orange": "The sampled ball is orange", "black": "The sampled ball is black", "green": "The sampled ball is green", "yellow": "The sampled ball is yellow", "pink": "The sampled ball is pink", "gray": "The sampled ball is gray", "brown": "The sampled ball is brown", "purple": "The sampled ball is purple"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a orange ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: brown, red, black, pink", "The drawn color is one of: purple, green, orange, gray", "The drawn color is one of: cyan, blue, white, yellow"]}}, "gold": {}, "gold_probs": {"outcome": {"red": 0.01818181818181818, "cyan": 0.05454545454545454, "blue": 0.09090909090909091, "white": 0.16363636363636364, "orange": 0.10909090909090909, "black": 0.10909090909090909, "green": 0.12727272727272726, "yellow": 0.03636363636363636, "pink": 0.03636363636363636, "gray": 0.05454545454545454, "brown": 0.12727272727272726, "purple": 0.07272727272727272}, "focus": {"false": 0.8909090909090909, "true": 0.10909090909090909}, "group": {"0": 0.2909090909090909, "1": 0.3636363636363636, "2": 0.34545454545454546}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"6/55\"],[\"blue\",\"1/11\"],[\"brown\",\"7/55\"],[\"cyan\",\"3/55\"],[\"gray\",\"3/55\"],[\"green\",\"7/55\"],[\"orange\",\"6/55\"],[\"pink\",\"2/55\"],[\"purple\",\"4/55\"],[\"red\",\"1/55\"],[\"white\",\"9/55\"],[\"yellow\",\"2/55\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"brown": 0.05, "yellow": 0, "cyan": 0, "white": 0.92, "gray": 0, "purple": 0, "black": 0, "green": 0.01, "blue": 0, "red": 0.02, "pink": 0, "orange": 0}, "focus": {"false": 0.86, "true": 0.14}, "group": {"0": 0.06, "1": 0.12, "2": 0.82}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:5", "state_id": "known_chance_v2:train:5", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: green=3, purple=5, brown=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"brown": "The sampled ball is brown", "green": "The sampled ball is green", "purple": "The sampled ball is purple"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a green ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: green", "The drawn color is one of: purple", "The drawn color is one of: brown"]}}, "gold": {}, "gold_probs": {"outcome": {"brown": 0.5, "green": 0.1875, "purple": 0.3125}, "focus": {"false": 0.8125, "true": 0.1875}, "group": {"0": 0.1875, "1": 0.3125, "2": 0.5}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"1/2\"],[\"green\",\"3/16\"],[\"purple\",\"5/16\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0.08, "brown": 0.9, "purple": 0.02}, "focus": {"false": 0.83, "true": 0.17}, "group": {"0": 0.02, "1": 0.02, "2": 0.96}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:6", "state_id": "known_chance_v2:train:6", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: brown=1, purple=2, gray=5, orange=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"purple": "The sampled ball is purple", "brown": "The sampled ball is brown", "orange": "The sampled ball is orange", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a brown ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: brown, orange", "The drawn color is one of: purple", "The drawn color is one of: gray"]}}, "gold": {}, "gold_probs": {"outcome": {"purple": 0.2222222222222222, "brown": 0.1111111111111111, "orange": 0.1111111111111111, "gray": 0.5555555555555556}, "focus": {"false": 0.8888888888888888, "true": 0.1111111111111111}, "group": {"0": 0.2222222222222222, "1": 0.2222222222222222, "2": 0.5555555555555556}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"1/9\"],[\"gray\",\"5/9\"],[\"orange\",\"1/9\"],[\"purple\",\"2/9\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.01, "brown": 0.02, "gray": 0.97, "orange": 0}, "focus": {"false": 0.9, "true": 0.1}, "group": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:7", "state_id": "known_chance_v2:train:7", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: red=4, pink=2, black=2, purple=4, brown=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"pink": "The sampled ball is pink", "black": "The sampled ball is black", "brown": "The sampled ball is brown", "red": "The sampled ball is red", "purple": "The sampled ball is purple"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a purple ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: red, purple", "The drawn color is one of: pink, brown", "The drawn color is one of: black"]}}, "gold": {}, "gold_probs": {"outcome": {"pink": 0.11764705882352941, "black": 0.11764705882352941, "brown": 0.29411764705882354, "red": 0.23529411764705882, "purple": 0.23529411764705882}, "focus": {"false": 0.7647058823529411, "true": 0.23529411764705882}, "group": {"0": 0.47058823529411764, "1": 0.4117647058823529, "2": 0.11764705882352941}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"2/17\"],[\"brown\",\"5/17\"],[\"pink\",\"2/17\"],[\"purple\",\"4/17\"],[\"red\",\"4/17\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0, "black": 0, "brown": 0.93, "purple": 0.01, "red": 0.06}, "focus": {"false": 0.77, "true": 0.23}, "group": {"0": 0.5, "1": 0.49, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:8", "state_id": "known_chance_v2:train:8", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: gray=2, yellow=5, blue=6, cyan=3, brown=2, black=5, white=7, red=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "blue": "The sampled ball is blue", "brown": "The sampled ball is brown", "yellow": "The sampled ball is yellow", "gray": "The sampled ball is gray", "cyan": "The sampled ball is cyan", "red": "The sampled ball is red", "white": "The sampled ball is white"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a red ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: gray, cyan, white", "The drawn color is one of: yellow, brown, red", "The drawn color is one of: blue, black"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.15151515151515152, "blue": 0.18181818181818182, "brown": 0.06060606060606061, "yellow": 0.15151515151515152, "gray": 0.06060606060606061, "cyan": 0.09090909090909091, "red": 0.09090909090909091, "white": 0.21212121212121213}, "focus": {"false": 0.9090909090909091, "true": 0.09090909090909091}, "group": {"0": 0.36363636363636365, "1": 0.30303030303030304, "2": 0.33333333333333337}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"5/33\"],[\"blue\",\"2/11\"],[\"brown\",\"2/33\"],[\"cyan\",\"1/11\"],[\"gray\",\"2/33\"],[\"red\",\"1/11\"],[\"white\",\"7/33\"],[\"yellow\",\"5/33\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"brown": 0, "gray": 0.03, "blue": 0.02, "white": 0.95, "yellow": 0, "red": 0, "cyan": 0, "black": 0}, "focus": {"false": 0.85, "true": 0.15}, "group": {"0": 0.27, "1": 0.04, "2": 0.69}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:9", "state_id": "known_chance_v2:train:9", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: green=7, red=4, white=5, purple=1, gray=5, cyan=2, blue=7, brown=9, orange=7, pink=8, black=5, yellow=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"purple": "The sampled ball is purple", "red": "The sampled ball is red", "cyan": "The sampled ball is cyan", "orange": "The sampled ball is orange", "blue": "The sampled ball is blue", "pink": "The sampled ball is pink", "yellow": "The sampled ball is yellow", "gray": "The sampled ball is gray", "green": "The sampled ball is green", "brown": "The sampled ball is brown", "white": "The sampled ball is white", "black": "The sampled ball is black"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a purple ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: green, purple, blue, pink", "The drawn color is one of: red, gray, brown, black", "The drawn color is one of: white, cyan, orange, yellow"]}}, "gold": {}, "gold_probs": {"outcome": {"purple": 0.014925373134328358, "red": 0.05970149253731343, "cyan": 0.029850746268656716, "orange": 0.1044776119402985, "blue": 0.1044776119402985, "pink": 0.11940298507462686, "yellow": 0.1044776119402985, "gray": 0.07462686567164178, "green": 0.1044776119402985, "brown": 0.13432835820895522, "white": 0.07462686567164178, "black": 0.07462686567164178}, "focus": {"false": 0.9850746268656716, "true": 0.014925373134328358}, "group": {"0": 0.34328358208955223, "1": 0.34328358208955223, "2": 0.3134328358208955}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"5/67\"],[\"blue\",\"7/67\"],[\"brown\",\"9/67\"],[\"cyan\",\"2/67\"],[\"gray\",\"5/67\"],[\"green\",\"7/67\"],[\"orange\",\"7/67\"],[\"pink\",\"8/67\"],[\"purple\",\"1/67\"],[\"red\",\"4/67\"],[\"white\",\"5/67\"],[\"yellow\",\"7/67\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.03, "black": 0, "cyan": 0, "brown": 0.83, "purple": 0.02, "green": 0.09999999999999999, "gray": 0, "blue": 0.01, "yellow": 0, "red": 0, "white": 0, "orange": 0.01}, "focus": {"false": 0.95, "true": 0.05}, "group": {"0": 0.32, "1": 0.56, "2": 0.12}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:13", "state_id": "known_chance_v2:train:13", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: red=2, black=5, blue=4, purple=6, brown=6, white=4, orange=1, green=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"brown": "The sampled ball is brown", "white": "The sampled ball is white", "purple": "The sampled ball is purple", "green": "The sampled ball is green", "blue": "The sampled ball is blue", "orange": "The sampled ball is orange", "black": "The sampled ball is black", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a orange ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: red, purple, orange", "The drawn color is one of: black, brown, green", "The drawn color is one of: blue, white"]}}, "gold": {}, "gold_probs": {"outcome": {"brown": 0.17142857142857143, "white": 0.11428571428571428, "purple": 0.17142857142857143, "green": 0.2, "blue": 0.11428571428571428, "orange": 0.02857142857142857, "black": 0.14285714285714285, "red": 0.05714285714285714}, "focus": {"false": 0.9714285714285714, "true": 0.02857142857142857}, "group": {"0": 0.2571428571428571, "1": 0.5142857142857142, "2": 0.22857142857142856}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/7\"],[\"blue\",\"4/35\"],[\"brown\",\"6/35\"],[\"green\",\"1/5\"],[\"orange\",\"1/35\"],[\"purple\",\"6/35\"],[\"red\",\"2/35\"],[\"white\",\"4/35\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0, "purple": 0.01, "red": 0.03, "black": 0.03, "orange": 0.03, "brown": 0.01, "blue": 0, "green": 0.89}, "focus": {"false": 0.94, "true": 0.06}, "group": {"0": 0, "1": 1, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:11", "state_id": "known_chance_v2:train:11", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: brown=6, orange=3, cyan=6, red=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "red": "The sampled ball is red", "orange": "The sampled ball is orange", "brown": "The sampled ball is brown"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a orange ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: brown, red", "The drawn color is one of: orange", "The drawn color is one of: cyan"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.2608695652173913, "red": 0.34782608695652173, "orange": 0.13043478260869565, "brown": 0.2608695652173913}, "focus": {"false": 0.8695652173913043, "true": 0.13043478260869565}, "group": {"0": 0.6086956521739131, "1": 0.13043478260869565, "2": 0.2608695652173913}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"6/23\"],[\"cyan\",\"6/23\"],[\"orange\",\"3/23\"],[\"red\",\"8/23\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"red": 0.93, "brown": 0.05, "cyan": 0.01, "orange": 0.01}, "focus": {"false": 0.83, "true": 0.17}, "group": {"0": 0.98, "1": 0.01, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:14", "state_id": "known_chance_v2:train:14", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: gray=2, blue=7, brown=1, black=6, orange=3, pink=6, green=1, yellow=6, white=2, red=6, purple=3, cyan=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"gray": "The sampled ball is gray", "blue": "The sampled ball is blue", "black": "The sampled ball is black", "yellow": "The sampled ball is yellow", "pink": "The sampled ball is pink", "orange": "The sampled ball is orange", "cyan": "The sampled ball is cyan", "white": "The sampled ball is white", "red": "The sampled ball is red", "purple": "The sampled ball is purple", "brown": "The sampled ball is brown", "green": "The sampled ball is green"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a green ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: gray, black, green, red", "The drawn color is one of: blue, orange, yellow, purple", "The drawn color is one of: brown, pink, white, cyan"]}}, "gold": {}, "gold_probs": {"outcome": {"gray": 0.043478260869565216, "blue": 0.15217391304347827, "black": 0.13043478260869565, "yellow": 0.13043478260869565, "pink": 0.13043478260869565, "orange": 0.06521739130434782, "cyan": 0.06521739130434782, "white": 0.043478260869565216, "red": 0.13043478260869565, "purple": 0.06521739130434782, "brown": 0.021739130434782608, "green": 0.021739130434782608}, "focus": {"false": 0.9782608695652174, "true": 0.021739130434782608}, "group": {"0": 0.32608695652173914, "1": 0.41304347826086957, "2": 0.2608695652173913}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"3/23\"],[\"blue\",\"7/46\"],[\"brown\",\"1/46\"],[\"cyan\",\"3/46\"],[\"gray\",\"1/23\"],[\"green\",\"1/46\"],[\"orange\",\"3/46\"],[\"pink\",\"3/23\"],[\"purple\",\"3/46\"],[\"red\",\"3/23\"],[\"white\",\"1/23\"],[\"yellow\",\"3/23\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.01, "gray": 0.01, "green": 0, "brown": 0.01, "blue": 0.93, "white": 0, "orange": 0, "yellow": 0.01, "purple": 0, "red": 0, "cyan": 0, "black": 0.02}, "focus": {"false": 0.95, "true": 0.05}, "group": {"0": 0.23, "1": 0.73, "2": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:12", "state_id": "known_chance_v2:train:12", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: blue=8, cyan=2, pink=4, red=3, black=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "cyan": "The sampled ball is cyan", "pink": "The sampled ball is pink", "red": "The sampled ball is red", "blue": "The sampled ball is blue"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue, red", "The drawn color is one of: cyan, black", "The drawn color is one of: pink"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.05555555555555555, "cyan": 0.1111111111111111, "pink": 0.2222222222222222, "red": 0.16666666666666666, "blue": 0.4444444444444444}, "focus": {"false": 0.8888888888888888, "true": 0.1111111111111111}, "group": {"0": 0.611111111111111, "1": 0.16666666666666666, "2": 0.2222222222222222}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/18\"],[\"blue\",\"4/9\"],[\"cyan\",\"1/9\"],[\"pink\",\"2/9\"],[\"red\",\"1/6\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"black": 0, "cyan": 0, "blue": 1, "red": 0, "pink": 0}, "focus": {"false": 0.89, "true": 0.11}, "group": {"0": 0.98, "1": 0.01, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:15", "state_id": "known_chance_v2:train:15", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: purple=2, cyan=9, gray=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"gray": "The sampled ball is gray", "cyan": "The sampled ball is cyan", "purple": "The sampled ball is purple"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: purple", "The drawn color is one of: cyan", "The drawn color is one of: gray"]}}, "gold": {}, "gold_probs": {"outcome": {"gray": 0.15384615384615385, "cyan": 0.6923076923076923, "purple": 0.15384615384615385}, "focus": {"false": 0.3076923076923077, "true": 0.6923076923076923}, "group": {"0": 0.15384615384615385, "1": 0.6923076923076923, "2": 0.15384615384615385}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"9/13\"],[\"gray\",\"2/13\"],[\"purple\",\"2/13\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0.99, "purple": 0.01, "gray": 0}, "focus": {"false": 0.35, "true": 0.65}, "group": {"0": 0, "1": 1, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:10", "state_id": "known_chance_v2:train:10", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: black=1, green=8, blue=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "green": "The sampled ball is green", "black": "The sampled ball is black"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a green ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: black", "The drawn color is one of: green", "The drawn color is one of: blue"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.47058823529411764, "green": 0.47058823529411764, "black": 0.058823529411764705}, "focus": {"false": 0.5294117647058824, "true": 0.47058823529411764}, "group": {"0": 0.058823529411764705, "1": 0.47058823529411764, "2": 0.47058823529411764}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/17\"],[\"blue\",\"8/17\"],[\"green\",\"8/17\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0.64, "black": 0.22, "blue": 0.14}, "focus": {"false": 0.64, "true": 0.36}, "group": {"0": 0.03, "1": 0.83, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:17", "state_id": "known_chance_v2:train:17", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: yellow=8, green=3, black=4, gray=1, brown=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"brown": "The sampled ball is brown", "gray": "The sampled ball is gray", "green": "The sampled ball is green", "black": "The sampled ball is black", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a brown ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: yellow, gray", "The drawn color is one of: green, brown", "The drawn color is one of: black"]}}, "gold": {}, "gold_probs": {"outcome": {"brown": 0.3333333333333333, "gray": 0.041666666666666664, "green": 0.125, "black": 0.16666666666666666, "yellow": 0.3333333333333333}, "focus": {"false": 0.6666666666666667, "true": 0.3333333333333333}, "group": {"0": 0.375, "1": 0.4583333333333333, "2": 0.16666666666666666}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/6\"],[\"brown\",\"1/3\"],[\"gray\",\"1/24\"],[\"green\",\"1/8\"],[\"yellow\",\"1/3\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"yellow": 0.91, "black": 0, "brown": 0.06, "green": 0, "gray": 0.03}, "focus": {"false": 0.78, "true": 0.22}, "group": {"0": 0.65, "1": 0.31, "2": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:16", "state_id": "known_chance_v2:train:16", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: orange=7, cyan=2, white=5, gray=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "gray": "The sampled ball is gray", "orange": "The sampled ball is orange", "white": "The sampled ball is white"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: orange, gray", "The drawn color is one of: cyan", "The drawn color is one of: white"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.125, "gray": 0.125, "orange": 0.4375, "white": 0.3125}, "focus": {"false": 0.875, "true": 0.125}, "group": {"0": 0.5625, "1": 0.125, "2": 0.3125}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"1/8\"],[\"gray\",\"1/8\"],[\"orange\",\"7/16\"],[\"white\",\"5/16\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"orange": 0.99, "gray": 0, "cyan": 0, "white": 0.01}, "focus": {"false": 0.84, "true": 0.16}, "group": {"0": 0.94, "1": 0.01, "2": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:19", "state_id": "known_chance_v2:train:19", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: green=9, brown=8, white=8, black=7, blue=9, yellow=4, pink=9, orange=2, gray=4, red=8, purple=8, cyan=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"yellow": "The sampled ball is yellow", "purple": "The sampled ball is purple", "gray": "The sampled ball is gray", "blue": "The sampled ball is blue", "green": "The sampled ball is green", "pink": "The sampled ball is pink", "red": "The sampled ball is red", "white": "The sampled ball is white", "cyan": "The sampled ball is cyan", "orange": "The sampled ball is orange", "black": "The sampled ball is black", "brown": "The sampled ball is brown"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a pink ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: green, black, pink, red", "The drawn color is one of: brown, blue, orange, purple", "The drawn color is one of: white, yellow, gray, cyan"]}}, "gold": {}, "gold_probs": {"outcome": {"yellow": 0.05, "purple": 0.1, "gray": 0.05, "blue": 0.1125, "green": 0.1125, "pink": 0.1125, "red": 0.1, "white": 0.1, "cyan": 0.05, "orange": 0.025, "black": 0.0875, "brown": 0.1}, "focus": {"false": 0.8875, "true": 0.1125}, "group": {"0": 0.4125, "1": 0.3375, "2": 0.25}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"7/80\"],[\"blue\",\"9/80\"],[\"brown\",\"1/10\"],[\"cyan\",\"1/20\"],[\"gray\",\"1/20\"],[\"green\",\"9/80\"],[\"orange\",\"1/40\"],[\"pink\",\"9/80\"],[\"purple\",\"1/10\"],[\"red\",\"1/10\"],[\"white\",\"1/10\"],[\"yellow\",\"1/20\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0.01, "orange": 0.1, "blue": 0.05, "yellow": 0.01, "brown": 0.02, "gray": 0, "black": 0.01, "pink": 0.06, "cyan": 0, "purple": 0.01, "green": 0.72, "red": 0.01}, "focus": {"false": 0.8200000000000001, "true": 0.18}, "group": {"0": 0.82, "1": 0.12, "2": 0.06}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:18", "state_id": "known_chance_v2:train:18", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: orange=7, black=8, brown=2, gray=1, pink=6, blue=5, purple=8, white=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"brown": "The sampled ball is brown", "purple": "The sampled ball is purple", "pink": "The sampled ball is pink", "white": "The sampled ball is white", "gray": "The sampled ball is gray", "blue": "The sampled ball is blue", "black": "The sampled ball is black", "orange": "The sampled ball is orange"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a white ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: orange, gray, purple", "The drawn color is one of: black, pink, white", "The drawn color is one of: brown, blue"]}}, "gold": {}, "gold_probs": {"outcome": {"brown": 0.046511627906976744, "purple": 0.18604651162790697, "pink": 0.13953488372093023, "white": 0.13953488372093023, "gray": 0.023255813953488372, "blue": 0.11627906976744186, "black": 0.18604651162790697, "orange": 0.16279069767441862}, "focus": {"false": 0.8604651162790697, "true": 0.13953488372093023}, "group": {"0": 0.37209302325581395, "1": 0.46511627906976744, "2": 0.16279069767441862}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"8/43\"],[\"blue\",\"5/43\"],[\"brown\",\"2/43\"],[\"gray\",\"1/43\"],[\"orange\",\"7/43\"],[\"pink\",\"6/43\"],[\"purple\",\"8/43\"],[\"white\",\"6/43\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"gray": 0.1, "purple": 0.11, "black": 0.61, "blue": 0.01, "white": 0.01, "orange": 0.14, "pink": 0.01, "brown": 0.01}, "focus": {"false": 0.85, "true": 0.15}, "group": {"0": 0.25, "1": 0.74, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:20", "state_id": "known_chance_v2:train:20", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: pink=9, gray=4, orange=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"gray": "The sampled ball is gray", "orange": "The sampled ball is orange", "pink": "The sampled ball is pink"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a orange ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: pink", "The drawn color is one of: gray", "The drawn color is one of: orange"]}}, "gold": {}, "gold_probs": {"outcome": {"gray": 0.25, "orange": 0.1875, "pink": 0.5625}, "focus": {"false": 0.8125, "true": 0.1875}, "group": {"0": 0.5625, "1": 0.25, "2": 0.1875}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"gray\",\"1/4\"],[\"orange\",\"3/16\"],[\"pink\",\"9/16\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 1, "gray": 0, "orange": 0}, "focus": {"false": 0.8200000000000001, "true": 0.18}, "group": {"0": 0.98, "1": 0.01, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:21", "state_id": "known_chance_v2:train:21", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: yellow=5, white=6, purple=6, cyan=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"yellow": "The sampled ball is yellow", "white": "The sampled ball is white", "cyan": "The sampled ball is cyan", "purple": "The sampled ball is purple"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: yellow, cyan", "The drawn color is one of: white", "The drawn color is one of: purple"]}}, "gold": {}, "gold_probs": {"outcome": {"yellow": 0.2777777777777778, "white": 0.3333333333333333, "cyan": 0.05555555555555555, "purple": 0.3333333333333333}, "focus": {"false": 0.7222222222222222, "true": 0.2777777777777778}, "group": {"0": 0.33333333333333337, "1": 0.3333333333333333, "2": 0.3333333333333333}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"1/18\"],[\"purple\",\"1/3\"],[\"white\",\"1/3\"],[\"yellow\",\"5/18\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0.44, "cyan": 0.07, "yellow": 0.45, "purple": 0.04}, "focus": {"false": 0.8, "true": 0.2}, "group": {"0": 0.14, "1": 0.58, "2": 0.28}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:24", "state_id": "known_chance_v2:train:24", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: yellow=9, purple=2, green=3, brown=1, pink=9, orange=8, blue=2, cyan=8, gray=2, red=1, white=2, black=9. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "orange": "The sampled ball is orange", "green": "The sampled ball is green", "brown": "The sampled ball is brown", "yellow": "The sampled ball is yellow", "black": "The sampled ball is black", "gray": "The sampled ball is gray", "purple": "The sampled ball is purple", "red": "The sampled ball is red", "pink": "The sampled ball is pink", "white": "The sampled ball is white", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a purple ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: yellow, brown, blue, red", "The drawn color is one of: purple, pink, cyan, white", "The drawn color is one of: green, orange, gray, black"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.03571428571428571, "orange": 0.14285714285714285, "green": 0.05357142857142857, "brown": 0.017857142857142856, "yellow": 0.16071428571428573, "black": 0.16071428571428573, "gray": 0.03571428571428571, "purple": 0.03571428571428571, "red": 0.017857142857142856, "pink": 0.16071428571428573, "white": 0.03571428571428571, "cyan": 0.14285714285714285}, "focus": {"false": 0.9642857142857143, "true": 0.03571428571428571}, "group": {"0": 0.23214285714285715, "1": 0.375, "2": 0.39285714285714285}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"9/56\"],[\"blue\",\"1/28\"],[\"brown\",\"1/56\"],[\"cyan\",\"1/7\"],[\"gray\",\"1/28\"],[\"green\",\"3/56\"],[\"orange\",\"1/7\"],[\"pink\",\"9/56\"],[\"purple\",\"1/28\"],[\"red\",\"1/56\"],[\"white\",\"1/28\"],[\"yellow\",\"9/56\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0, "pink": 0.05, "green": 0.01, "orange": 0.02, "blue": 0, "red": 0.01, "cyan": 0.01, "white": 0, "black": 0.14, "brown": 0.01, "yellow": 0.75, "gray": 0}, "focus": {"false": 0.91, "true": 0.09}, "group": {"0": 0.23, "1": 0.27, "2": 0.5}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:25", "state_id": "known_chance_v2:train:25", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: gray=6, yellow=3, purple=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"purple": "The sampled ball is purple", "gray": "The sampled ball is gray", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a purple ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: gray", "The drawn color is one of: yellow", "The drawn color is one of: purple"]}}, "gold": {}, "gold_probs": {"outcome": {"purple": 0.47058823529411764, "gray": 0.35294117647058826, "yellow": 0.17647058823529413}, "focus": {"false": 0.5294117647058824, "true": 0.47058823529411764}, "group": {"0": 0.35294117647058826, "1": 0.17647058823529413, "2": 0.47058823529411764}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"gray\",\"6/17\"],[\"purple\",\"8/17\"],[\"yellow\",\"3/17\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.77, "yellow": 0.01, "gray": 0.22}, "focus": {"false": 0.6799999999999999, "true": 0.32}, "group": {"0": 0.06, "1": 0.01, "2": 0.93}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:22", "state_id": "known_chance_v2:train:22", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: yellow=5, green=8, cyan=7, gray=1, purple=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "purple": "The sampled ball is purple", "gray": "The sampled ball is gray", "yellow": "The sampled ball is yellow", "green": "The sampled ball is green"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: yellow, gray", "The drawn color is one of: green, purple", "The drawn color is one of: cyan"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.2916666666666667, "purple": 0.125, "gray": 0.041666666666666664, "yellow": 0.20833333333333334, "green": 0.3333333333333333}, "focus": {"false": 0.7916666666666666, "true": 0.20833333333333334}, "group": {"0": 0.25, "1": 0.4583333333333333, "2": 0.2916666666666667}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"7/24\"],[\"gray\",\"1/24\"],[\"green\",\"1/3\"],[\"purple\",\"1/8\"],[\"yellow\",\"5/24\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0, "cyan": 0.01, "gray": 0.06, "green": 0.89, "yellow": 0.04}, "focus": {"false": 0.83, "true": 0.17}, "group": {"0": 0.03, "1": 0.89, "2": 0.08}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:26", "state_id": "known_chance_v2:train:26", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: cyan=5, white=3, yellow=7, pink=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "yellow": "The sampled ball is yellow", "pink": "The sampled ball is pink", "white": "The sampled ball is white"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a white ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: cyan, pink", "The drawn color is one of: white", "The drawn color is one of: yellow"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.23809523809523808, "yellow": 0.3333333333333333, "pink": 0.2857142857142857, "white": 0.14285714285714285}, "focus": {"false": 0.8571428571428572, "true": 0.14285714285714285}, "group": {"0": 0.5238095238095237, "1": 0.14285714285714285, "2": 0.3333333333333333}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"5/21\"],[\"pink\",\"2/7\"],[\"white\",\"1/7\"],[\"yellow\",\"1/3\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0.12, "white": 0.01, "pink": 0.01, "yellow": 0.86}, "focus": {"false": 0.81, "true": 0.19}, "group": {"0": 0.1, "1": 0.01, "2": 0.89}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:28", "state_id": "known_chance_v2:train:28", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: brown=2, cyan=7, green=7, black=4, purple=3, gray=2, orange=4, white=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"gray": "The sampled ball is gray", "green": "The sampled ball is green", "cyan": "The sampled ball is cyan", "purple": "The sampled ball is purple", "white": "The sampled ball is white", "brown": "The sampled ball is brown", "orange": "The sampled ball is orange", "black": "The sampled ball is black"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a purple ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: brown, black, orange", "The drawn color is one of: cyan, purple, white", "The drawn color is one of: green, gray"]}}, "gold": {}, "gold_probs": {"outcome": {"gray": 0.05555555555555555, "green": 0.19444444444444445, "cyan": 0.19444444444444445, "purple": 0.08333333333333333, "white": 0.19444444444444445, "brown": 0.05555555555555555, "orange": 0.1111111111111111, "black": 0.1111111111111111}, "focus": {"false": 0.9166666666666666, "true": 0.08333333333333333}, "group": {"0": 0.2777777777777778, "1": 0.4722222222222222, "2": 0.25}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/9\"],[\"brown\",\"1/18\"],[\"cyan\",\"7/36\"],[\"gray\",\"1/18\"],[\"green\",\"7/36\"],[\"orange\",\"1/9\"],[\"purple\",\"1/12\"],[\"white\",\"7/36\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0.16, "orange": 0.01, "brown": 0.09, "gray": 0.01, "black": 0.01, "cyan": 0.58, "white": 0.13, "purple": 0.01}, "focus": {"false": 0.87, "true": 0.13}, "group": {"0": 0.05, "1": 0.93, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:23", "state_id": "known_chance_v2:train:23", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: red=2, black=8, yellow=2, purple=3, green=7, orange=2, white=1, cyan=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"white": "The sampled ball is white", "green": "The sampled ball is green", "purple": "The sampled ball is purple", "black": "The sampled ball is black", "cyan": "The sampled ball is cyan", "red": "The sampled ball is red", "yellow": "The sampled ball is yellow", "orange": "The sampled ball is orange"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: red, purple, white", "The drawn color is one of: black, green, cyan", "The drawn color is one of: yellow, orange"]}}, "gold": {}, "gold_probs": {"outcome": {"white": 0.03571428571428571, "green": 0.25, "purple": 0.10714285714285714, "black": 0.2857142857142857, "cyan": 0.10714285714285714, "red": 0.07142857142857142, "yellow": 0.07142857142857142, "orange": 0.07142857142857142}, "focus": {"false": 0.7142857142857143, "true": 0.2857142857142857}, "group": {"0": 0.21428571428571427, "1": 0.6428571428571428, "2": 0.14285714285714285}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"2/7\"],[\"cyan\",\"3/28\"],[\"green\",\"1/4\"],[\"orange\",\"1/14\"],[\"purple\",\"3/28\"],[\"red\",\"1/14\"],[\"white\",\"1/28\"],[\"yellow\",\"1/14\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"orange": 0, "white": 0.01, "purple": 0, "red": 0.01, "cyan": 0, "black": 0.95, "yellow": 0, "green": 0.03}, "focus": {"false": 0.74, "true": 0.26}, "group": {"0": 0, "1": 1, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:27", "state_id": "known_chance_v2:train:27", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: cyan=2, purple=9, blue=9, orange=7, red=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "cyan": "The sampled ball is cyan", "purple": "The sampled ball is purple", "orange": "The sampled ball is orange", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: cyan, orange", "The drawn color is one of: purple, red", "The drawn color is one of: blue"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.2903225806451613, "cyan": 0.06451612903225806, "purple": 0.2903225806451613, "orange": 0.22580645161290322, "red": 0.12903225806451613}, "focus": {"false": 0.935483870967742, "true": 0.06451612903225806}, "group": {"0": 0.29032258064516125, "1": 0.41935483870967744, "2": 0.2903225806451613}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"9/31\"],[\"cyan\",\"2/31\"],[\"orange\",\"7/31\"],[\"purple\",\"9/31\"],[\"red\",\"4/31\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"orange": 0.02, "purple": 0.8, "blue": 0.09, "cyan": 0.08, "red": 0.01}, "focus": {"false": 0.91, "true": 0.09}, "group": {"0": 0.04, "1": 0.72, "2": 0.24}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:29", "state_id": "known_chance_v2:train:29", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: white=4, orange=5, cyan=4, blue=7, purple=3, black=8, brown=2, green=8, pink=1, gray=2, yellow=2, red=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"red": "The sampled ball is red", "white": "The sampled ball is white", "gray": "The sampled ball is gray", "black": "The sampled ball is black", "orange": "The sampled ball is orange", "brown": "The sampled ball is brown", "purple": "The sampled ball is purple", "blue": "The sampled ball is blue", "yellow": "The sampled ball is yellow", "green": "The sampled ball is green", "pink": "The sampled ball is pink", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a pink ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white, blue, brown, gray", "The drawn color is one of: orange, purple, green, yellow", "The drawn color is one of: cyan, black, pink, red"]}}, "gold": {}, "gold_probs": {"outcome": {"red": 0.061224489795918366, "white": 0.08163265306122448, "gray": 0.04081632653061224, "black": 0.16326530612244897, "orange": 0.10204081632653061, "brown": 0.04081632653061224, "purple": 0.061224489795918366, "blue": 0.14285714285714285, "yellow": 0.04081632653061224, "green": 0.16326530612244897, "pink": 0.02040816326530612, "cyan": 0.08163265306122448}, "focus": {"false": 0.9795918367346939, "true": 0.02040816326530612}, "group": {"0": 0.30612244897959184, "1": 0.36734693877551017, "2": 0.32653061224489793}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"8/49\"],[\"blue\",\"1/7\"],[\"brown\",\"2/49\"],[\"cyan\",\"4/49\"],[\"gray\",\"2/49\"],[\"green\",\"8/49\"],[\"orange\",\"5/49\"],[\"pink\",\"1/49\"],[\"purple\",\"3/49\"],[\"red\",\"3/49\"],[\"white\",\"4/49\"],[\"yellow\",\"2/49\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0.04, "purple": 0, "yellow": 0, "pink": 0.05, "gray": 0, "red": 0.01, "black": 0.51, "cyan": 0.01, "orange": 0.02, "brown": 0.01, "blue": 0.15, "green": 0.2}, "focus": {"false": 0.95, "true": 0.05}, "group": {"0": 0.23, "1": 0.23, "2": 0.54}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:31", "state_id": "known_chance_v2:train:31", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: orange=6, red=3, black=9, cyan=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "red": "The sampled ball is red", "cyan": "The sampled ball is cyan", "orange": "The sampled ball is orange"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: orange, cyan", "The drawn color is one of: red", "The drawn color is one of: black"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.375, "red": 0.125, "cyan": 0.25, "orange": 0.25}, "focus": {"false": 0.625, "true": 0.375}, "group": {"0": 0.5, "1": 0.125, "2": 0.375}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"3/8\"],[\"cyan\",\"1/4\"],[\"orange\",\"1/4\"],[\"red\",\"1/8\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0.01, "red": 0.01, "orange": 0.05, "black": 0.93}, "focus": {"false": 0.69, "true": 0.31}, "group": {"0": 0.05, "1": 0, "2": 0.95}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:32", "state_id": "known_chance_v2:train:32", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: red=4, brown=8, yellow=8, white=1, orange=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"red": "The sampled ball is red", "yellow": "The sampled ball is yellow", "white": "The sampled ball is white", "orange": "The sampled ball is orange", "brown": "The sampled ball is brown"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a orange ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: red, white", "The drawn color is one of: brown, orange", "The drawn color is one of: yellow"]}}, "gold": {}, "gold_probs": {"outcome": {"red": 0.15384615384615385, "yellow": 0.3076923076923077, "white": 0.038461538461538464, "orange": 0.19230769230769232, "brown": 0.3076923076923077}, "focus": {"false": 0.8076923076923077, "true": 0.19230769230769232}, "group": {"0": 0.19230769230769232, "1": 0.5, "2": 0.3076923076923077}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"4/13\"],[\"orange\",\"5/26\"],[\"red\",\"2/13\"],[\"white\",\"1/26\"],[\"yellow\",\"4/13\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"yellow": 0.11, "brown": 0.77, "red": 0.05, "white": 0.06, "orange": 0.01}, "focus": {"false": 0.8200000000000001, "true": 0.18}, "group": {"0": 0.02, "1": 0.84, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:30", "state_id": "known_chance_v2:train:30", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: black=7, blue=9, red=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "red": "The sampled ball is red", "black": "The sampled ball is black"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: black", "The drawn color is one of: blue", "The drawn color is one of: red"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.47368421052631576, "red": 0.15789473684210525, "black": 0.3684210526315789}, "focus": {"false": 0.5263157894736843, "true": 0.47368421052631576}, "group": {"0": 0.3684210526315789, "1": 0.47368421052631576, "2": 0.15789473684210525}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"7/19\"],[\"blue\",\"9/19\"],[\"red\",\"3/19\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"red": 0.01, "blue": 0.74, "black": 0.25}, "focus": {"false": 0.65, "true": 0.35}, "group": {"0": 0.04, "1": 0.95, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:33", "state_id": "known_chance_v2:train:33", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: white=8, purple=4, red=7, blue=9, cyan=1, black=5, yellow=5, green=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "black": "The sampled ball is black", "green": "The sampled ball is green", "yellow": "The sampled ball is yellow", "purple": "The sampled ball is purple", "blue": "The sampled ball is blue", "red": "The sampled ball is red", "white": "The sampled ball is white"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a green ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white, blue, yellow", "The drawn color is one of: purple, cyan, green", "The drawn color is one of: red, black"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.02127659574468085, "black": 0.10638297872340426, "green": 0.1702127659574468, "yellow": 0.10638297872340426, "purple": 0.0851063829787234, "blue": 0.19148936170212766, "red": 0.14893617021276595, "white": 0.1702127659574468}, "focus": {"false": 0.8297872340425532, "true": 0.1702127659574468}, "group": {"0": 0.46808510638297873, "1": 0.2765957446808511, "2": 0.2553191489361702}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"5/47\"],[\"blue\",\"9/47\"],[\"cyan\",\"1/47\"],[\"green\",\"8/47\"],[\"purple\",\"4/47\"],[\"red\",\"7/47\"],[\"white\",\"8/47\"],[\"yellow\",\"5/47\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0.09, "cyan": 0.03, "black": 0, "green": 0.01, "yellow": 0, "blue": 0.87, "red": 0, "purple": 0}, "focus": {"false": 0.86, "true": 0.14}, "group": {"0": 0.97, "1": 0.01, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:35", "state_id": "known_chance_v2:train:35", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: black=8, white=1, cyan=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"white": "The sampled ball is white", "black": "The sampled ball is black", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: black", "The drawn color is one of: white", "The drawn color is one of: cyan"]}}, "gold": {}, "gold_probs": {"outcome": {"white": 0.058823529411764705, "black": 0.47058823529411764, "cyan": 0.47058823529411764}, "focus": {"false": 0.5294117647058824, "true": 0.47058823529411764}, "group": {"0": 0.47058823529411764, "1": 0.058823529411764705, "2": 0.47058823529411764}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"8/17\"],[\"cyan\",\"8/17\"],[\"white\",\"1/17\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"black": 0.92, "cyan": 0.05, "white": 0.03}, "focus": {"false": 0.61, "true": 0.39}, "group": {"0": 0.85, "1": 0.03, "2": 0.12}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:36", "state_id": "known_chance_v2:train:36", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: red=3, pink=1, blue=4, yellow=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"pink": "The sampled ball is pink", "yellow": "The sampled ball is yellow", "blue": "The sampled ball is blue", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a pink ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: red, yellow", "The drawn color is one of: pink", "The drawn color is one of: blue"]}}, "gold": {}, "gold_probs": {"outcome": {"pink": 0.07692307692307693, "yellow": 0.38461538461538464, "blue": 0.3076923076923077, "red": 0.23076923076923078}, "focus": {"false": 0.9230769230769231, "true": 0.07692307692307693}, "group": {"0": 0.6153846153846154, "1": 0.07692307692307693, "2": 0.3076923076923077}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"4/13\"],[\"pink\",\"1/13\"],[\"red\",\"3/13\"],[\"yellow\",\"5/13\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"yellow": 0.89, "blue": 0.04, "red": 0.06, "pink": 0.01}, "focus": {"false": 0.89, "true": 0.11}, "group": {"0": 0.87, "1": 0.01, "2": 0.12}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:34", "state_id": "known_chance_v2:train:34", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: brown=3, red=7, purple=2, pink=6, blue=2, gray=5, orange=5, yellow=9, green=6, cyan=3, black=8, white=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"brown": "The sampled ball is brown", "green": "The sampled ball is green", "pink": "The sampled ball is pink", "purple": "The sampled ball is purple", "orange": "The sampled ball is orange", "white": "The sampled ball is white", "blue": "The sampled ball is blue", "yellow": "The sampled ball is yellow", "black": "The sampled ball is black", "red": "The sampled ball is red", "cyan": "The sampled ball is cyan", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a red ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: brown, pink, orange, cyan", "The drawn color is one of: red, blue, yellow, black", "The drawn color is one of: purple, gray, green, white"]}}, "gold": {}, "gold_probs": {"outcome": {"brown": 0.05263157894736842, "green": 0.10526315789473684, "pink": 0.10526315789473684, "purple": 0.03508771929824561, "orange": 0.08771929824561403, "white": 0.017543859649122806, "blue": 0.03508771929824561, "yellow": 0.15789473684210525, "black": 0.14035087719298245, "red": 0.12280701754385964, "cyan": 0.05263157894736842, "gray": 0.08771929824561403}, "focus": {"false": 0.8771929824561404, "true": 0.12280701754385964}, "group": {"0": 0.2982456140350877, "1": 0.45614035087719296, "2": 0.24561403508771928}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"8/57\"],[\"blue\",\"2/57\"],[\"brown\",\"1/19\"],[\"cyan\",\"1/19\"],[\"gray\",\"5/57\"],[\"green\",\"2/19\"],[\"orange\",\"5/57\"],[\"pink\",\"2/19\"],[\"purple\",\"2/57\"],[\"red\",\"7/57\"],[\"white\",\"1/57\"],[\"yellow\",\"3/19\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0, "purple": 0, "brown": 0.03, "orange": 0, "black": 0.02, "pink": 0, "yellow": 0.87, "gray": 0, "cyan": 0, "blue": 0, "white": 0.05, "red": 0.03}, "focus": {"false": 0.81, "true": 0.19}, "group": {"0": 0.01, "1": 0.98, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:37", "state_id": "known_chance_v2:train:37", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: gray=1, pink=5, purple=6, blue=3, green=9. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"pink": "The sampled ball is pink", "purple": "The sampled ball is purple", "blue": "The sampled ball is blue", "gray": "The sampled ball is gray", "green": "The sampled ball is green"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a gray ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: gray, blue", "The drawn color is one of: pink, green", "The drawn color is one of: purple"]}}, "gold": {}, "gold_probs": {"outcome": {"pink": 0.20833333333333334, "purple": 0.25, "blue": 0.125, "gray": 0.041666666666666664, "green": 0.375}, "focus": {"false": 0.9583333333333334, "true": 0.041666666666666664}, "group": {"0": 0.16666666666666666, "1": 0.5833333333333334, "2": 0.25}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"1/8\"],[\"gray\",\"1/24\"],[\"green\",\"3/8\"],[\"pink\",\"5/24\"],[\"purple\",\"1/4\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.01, "gray": 0.06, "green": 0.9, "blue": 0, "purple": 0.03}, "focus": {"false": 0.94, "true": 0.06}, "group": {"0": 0.01, "1": 0.96, "2": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:40", "state_id": "known_chance_v2:train:40", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: pink=1, black=6, cyan=9. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "pink": "The sampled ball is pink", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a pink ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: pink", "The drawn color is one of: black", "The drawn color is one of: cyan"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.375, "pink": 0.0625, "cyan": 0.5625}, "focus": {"false": 0.9375, "true": 0.0625}, "group": {"0": 0.0625, "1": 0.375, "2": 0.5625}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"3/8\"],[\"cyan\",\"9/16\"],[\"pink\",\"1/16\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"black": 0.17, "cyan": 0.75, "pink": 0.08}, "focus": {"false": 0.9299999999999999, "true": 0.07}, "group": {"0": 0.01, "1": 0.07, "2": 0.92}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:38", "state_id": "known_chance_v2:train:38", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: purple=8, blue=9, black=6, cyan=6, green=2, red=6, white=8, yellow=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"yellow": "The sampled ball is yellow", "cyan": "The sampled ball is cyan", "black": "The sampled ball is black", "white": "The sampled ball is white", "blue": "The sampled ball is blue", "purple": "The sampled ball is purple", "green": "The sampled ball is green", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: purple, cyan, white", "The drawn color is one of: blue, green, yellow", "The drawn color is one of: black, red"]}}, "gold": {}, "gold_probs": {"outcome": {"yellow": 0.1, "cyan": 0.12, "black": 0.12, "white": 0.16, "blue": 0.18, "purple": 0.16, "green": 0.04, "red": 0.12}, "focus": {"false": 0.88, "true": 0.12}, "group": {"0": 0.44, "1": 0.32, "2": 0.24}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"3/25\"],[\"blue\",\"9/50\"],[\"cyan\",\"3/25\"],[\"green\",\"1/25\"],[\"purple\",\"4/25\"],[\"red\",\"3/25\"],[\"white\",\"4/25\"],[\"yellow\",\"1/10\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.03, "green": 0.05, "blue": 0.9, "cyan": 0, "red": 0, "white": 0.01, "black": 0.01, "yellow": 0}, "focus": {"false": 0.83, "true": 0.17}, "group": {"0": 0.71, "1": 0.19, "2": 0.1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:39", "state_id": "known_chance_v2:train:39", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: orange=1, blue=3, brown=1, red=5, green=6, cyan=5, black=9, purple=2, gray=8, white=8, pink=8, yellow=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "pink": "The sampled ball is pink", "cyan": "The sampled ball is cyan", "gray": "The sampled ball is gray", "orange": "The sampled ball is orange", "blue": "The sampled ball is blue", "white": "The sampled ball is white", "brown": "The sampled ball is brown", "purple": "The sampled ball is purple", "red": "The sampled ball is red", "yellow": "The sampled ball is yellow", "black": "The sampled ball is black"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a green ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: orange, red, black, white", "The drawn color is one of: blue, green, purple, pink", "The drawn color is one of: brown, cyan, gray, yellow"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.1016949152542373, "pink": 0.13559322033898305, "cyan": 0.0847457627118644, "gray": 0.13559322033898305, "orange": 0.01694915254237288, "blue": 0.05084745762711865, "white": 0.13559322033898305, "brown": 0.01694915254237288, "purple": 0.03389830508474576, "red": 0.0847457627118644, "yellow": 0.05084745762711865, "black": 0.15254237288135594}, "focus": {"false": 0.8983050847457628, "true": 0.1016949152542373}, "group": {"0": 0.3898305084745763, "1": 0.3220338983050848, "2": 0.288135593220339}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"9/59\"],[\"blue\",\"3/59\"],[\"brown\",\"1/59\"],[\"cyan\",\"5/59\"],[\"gray\",\"8/59\"],[\"green\",\"6/59\"],[\"orange\",\"1/59\"],[\"pink\",\"8/59\"],[\"purple\",\"2/59\"],[\"red\",\"5/59\"],[\"white\",\"8/59\"],[\"yellow\",\"3/59\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"brown": 0.01, "blue": 0, "white": 0, "gray": 0, "yellow": 0, "purple": 0, "red": 0, "orange": 0.060000000000000005, "pink": 0, "cyan": 0, "green": 0.02, "black": 0.91}, "focus": {"false": 0.86, "true": 0.14}, "group": {"0": 0.88, "1": 0.06, "2": 0.06}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:41", "state_id": "known_chance_v2:train:41", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: cyan=1, white=6, purple=8, gray=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"white": "The sampled ball is white", "gray": "The sampled ball is gray", "purple": "The sampled ball is purple", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a gray ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: cyan, gray", "The drawn color is one of: white", "The drawn color is one of: purple"]}}, "gold": {}, "gold_probs": {"outcome": {"white": 0.2727272727272727, "gray": 0.3181818181818182, "purple": 0.36363636363636365, "cyan": 0.045454545454545456}, "focus": {"false": 0.6818181818181819, "true": 0.3181818181818182}, "group": {"0": 0.36363636363636365, "1": 0.2727272727272727, "2": 0.36363636363636365}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"1/22\"],[\"gray\",\"7/22\"],[\"purple\",\"4/11\"],[\"white\",\"3/11\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"gray": 0.01, "white": 0.03, "purple": 0.9, "cyan": 0.06}, "focus": {"false": 0.76, "true": 0.24}, "group": {"0": 0.01, "1": 0.02, "2": 0.97}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:42", "state_id": "known_chance_v2:train:42", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: white=5, gray=5, brown=3, purple=4, yellow=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"yellow": "The sampled ball is yellow", "gray": "The sampled ball is gray", "brown": "The sampled ball is brown", "white": "The sampled ball is white", "purple": "The sampled ball is purple"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a purple ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white, purple", "The drawn color is one of: gray, yellow", "The drawn color is one of: brown"]}}, "gold": {}, "gold_probs": {"outcome": {"yellow": 0.19047619047619047, "gray": 0.23809523809523808, "brown": 0.14285714285714285, "white": 0.23809523809523808, "purple": 0.19047619047619047}, "focus": {"false": 0.8095238095238095, "true": 0.19047619047619047}, "group": {"0": 0.42857142857142855, "1": 0.42857142857142855, "2": 0.14285714285714285}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"1/7\"],[\"gray\",\"5/21\"],[\"purple\",\"4/21\"],[\"white\",\"5/21\"],[\"yellow\",\"4/21\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"yellow": 0.01, "brown": 0.03, "purple": 0.01, "white": 0.88, "gray": 0.07}, "focus": {"false": 0.81, "true": 0.19}, "group": {"0": 0.79, "1": 0.16, "2": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:43", "state_id": "known_chance_v2:train:43", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: pink=6, white=4, green=4, brown=2, blue=3, purple=4, orange=5, gray=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "purple": "The sampled ball is purple", "gray": "The sampled ball is gray", "white": "The sampled ball is white", "pink": "The sampled ball is pink", "orange": "The sampled ball is orange", "brown": "The sampled ball is brown", "blue": "The sampled ball is blue"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a brown ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: pink, brown, orange", "The drawn color is one of: white, blue, gray", "The drawn color is one of: green, purple"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.12121212121212122, "purple": 0.12121212121212122, "gray": 0.15151515151515152, "white": 0.12121212121212122, "pink": 0.18181818181818182, "orange": 0.15151515151515152, "brown": 0.06060606060606061, "blue": 0.09090909090909091}, "focus": {"false": 0.9393939393939394, "true": 0.06060606060606061}, "group": {"0": 0.3939393939393939, "1": 0.36363636363636365, "2": 0.24242424242424243}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"1/11\"],[\"brown\",\"2/33\"],[\"gray\",\"5/33\"],[\"green\",\"4/33\"],[\"orange\",\"5/33\"],[\"pink\",\"2/11\"],[\"purple\",\"4/33\"],[\"white\",\"4/33\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.98, "brown": 0.01, "purple": 0, "green": 0, "gray": 0, "blue": 0, "white": 0, "orange": 0.01}, "focus": {"false": 0.9, "true": 0.1}, "group": {"0": 0.82, "1": 0.11, "2": 0.07}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:45", "state_id": "known_chance_v2:train:45", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: yellow=3, cyan=3, gray=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"gray": "The sampled ball is gray", "yellow": "The sampled ball is yellow", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: yellow", "The drawn color is one of: cyan", "The drawn color is one of: gray"]}}, "gold": {}, "gold_probs": {"outcome": {"gray": 0.45454545454545453, "yellow": 0.2727272727272727, "cyan": 0.2727272727272727}, "focus": {"false": 0.7272727272727273, "true": 0.2727272727272727}, "group": {"0": 0.2727272727272727, "1": 0.2727272727272727, "2": 0.45454545454545453}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"3/11\"],[\"gray\",\"5/11\"],[\"yellow\",\"3/11\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"yellow": 0.07, "gray": 0.92, "cyan": 0.01}, "focus": {"false": 0.74, "true": 0.26}, "group": {"0": 0.02, "1": 0.01, "2": 0.97}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:44", "state_id": "known_chance_v2:train:44", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: cyan=7, orange=4, yellow=1, pink=3, red=3, blue=4, black=9, purple=6, green=9, brown=3, white=6, gray=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "pink": "The sampled ball is pink", "white": "The sampled ball is white", "brown": "The sampled ball is brown", "cyan": "The sampled ball is cyan", "yellow": "The sampled ball is yellow", "blue": "The sampled ball is blue", "gray": "The sampled ball is gray", "red": "The sampled ball is red", "orange": "The sampled ball is orange", "green": "The sampled ball is green", "purple": "The sampled ball is purple"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: cyan, pink, black, brown", "The drawn color is one of: orange, red, purple, white", "The drawn color is one of: yellow, blue, green, gray"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.15254237288135594, "pink": 0.05084745762711865, "white": 0.1016949152542373, "brown": 0.05084745762711865, "cyan": 0.11864406779661017, "yellow": 0.01694915254237288, "blue": 0.06779661016949153, "gray": 0.06779661016949153, "red": 0.05084745762711865, "orange": 0.06779661016949153, "green": 0.15254237288135594, "purple": 0.1016949152542373}, "focus": {"false": 0.9322033898305084, "true": 0.06779661016949153}, "group": {"0": 0.3728813559322034, "1": 0.3220338983050848, "2": 0.3050847457627119}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"9/59\"],[\"blue\",\"4/59\"],[\"brown\",\"3/59\"],[\"cyan\",\"7/59\"],[\"gray\",\"4/59\"],[\"green\",\"9/59\"],[\"orange\",\"4/59\"],[\"pink\",\"3/59\"],[\"purple\",\"6/59\"],[\"red\",\"3/59\"],[\"white\",\"6/59\"],[\"yellow\",\"1/59\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.01, "red": 0.01, "yellow": 0.13, "white": 0.01, "blue": 0.01, "pink": 0, "cyan": 0.19999999999999998, "orange": 0.01, "gray": 0.01, "green": 0.14, "black": 0.45999999999999996, "brown": 0.01}, "focus": {"false": 0.86, "true": 0.14}, "group": {"0": 0.78, "1": 0.1, "2": 0.12}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:46", "state_id": "known_chance_v2:train:46", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: orange=4, red=8, cyan=7, black=9. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"orange": "The sampled ball is orange", "cyan": "The sampled ball is cyan", "red": "The sampled ball is red", "black": "The sampled ball is black"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a red ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: orange, black", "The drawn color is one of: red", "The drawn color is one of: cyan"]}}, "gold": {}, "gold_probs": {"outcome": {"orange": 0.14285714285714285, "cyan": 0.25, "red": 0.2857142857142857, "black": 0.32142857142857145}, "focus": {"false": 0.7142857142857143, "true": 0.2857142857142857}, "group": {"0": 0.4642857142857143, "1": 0.2857142857142857, "2": 0.25}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"9/28\"],[\"cyan\",\"1/4\"],[\"orange\",\"1/7\"],[\"red\",\"2/7\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0.01, "orange": 0.13, "red": 0.12, "black": 0.74}, "focus": {"false": 0.74, "true": 0.26}, "group": {"0": 0.59, "1": 0.36, "2": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:48", "state_id": "known_chance_v2:train:48", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: black=3, yellow=5, pink=5, cyan=3, red=3, purple=9, white=6, brown=9. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"yellow": "The sampled ball is yellow", "white": "The sampled ball is white", "red": "The sampled ball is red", "pink": "The sampled ball is pink", "brown": "The sampled ball is brown", "purple": "The sampled ball is purple", "black": "The sampled ball is black", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a brown ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: black, cyan, white", "The drawn color is one of: yellow, red, brown", "The drawn color is one of: pink, purple"]}}, "gold": {}, "gold_probs": {"outcome": {"yellow": 0.11627906976744186, "white": 0.13953488372093023, "red": 0.06976744186046512, "pink": 0.11627906976744186, "brown": 0.20930232558139536, "purple": 0.20930232558139536, "black": 0.06976744186046512, "cyan": 0.06976744186046512}, "focus": {"false": 0.7906976744186046, "true": 0.20930232558139536}, "group": {"0": 0.27906976744186046, "1": 0.39534883720930236, "2": 0.32558139534883723}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"3/43\"],[\"brown\",\"9/43\"],[\"cyan\",\"3/43\"],[\"pink\",\"5/43\"],[\"purple\",\"9/43\"],[\"red\",\"3/43\"],[\"white\",\"6/43\"],[\"yellow\",\"5/43\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.01, "purple": 0.7, "yellow": 0.01, "white": 0.01, "cyan": 0.01, "red": 0, "black": 0.04, "brown": 0.22}, "focus": {"false": 0.8200000000000001, "true": 0.18}, "group": {"0": 0.02, "1": 0.17, "2": 0.8099999999999999}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:49", "state_id": "known_chance_v2:train:49", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: pink=8, black=6, red=2, yellow=1, blue=7, gray=1, orange=8, white=1, green=3, brown=1, cyan=2, purple=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"orange": "The sampled ball is orange", "blue": "The sampled ball is blue", "green": "The sampled ball is green", "red": "The sampled ball is red", "pink": "The sampled ball is pink", "cyan": "The sampled ball is cyan", "gray": "The sampled ball is gray", "white": "The sampled ball is white", "brown": "The sampled ball is brown", "purple": "The sampled ball is purple", "black": "The sampled ball is black", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a green ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: pink, yellow, orange, brown", "The drawn color is one of: black, blue, white, cyan", "The drawn color is one of: red, gray, green, purple"]}}, "gold": {}, "gold_probs": {"outcome": {"orange": 0.17777777777777778, "blue": 0.15555555555555556, "green": 0.06666666666666667, "red": 0.044444444444444446, "pink": 0.17777777777777778, "cyan": 0.044444444444444446, "gray": 0.022222222222222223, "white": 0.022222222222222223, "brown": 0.022222222222222223, "purple": 0.1111111111111111, "black": 0.13333333333333333, "yellow": 0.022222222222222223}, "focus": {"false": 0.9333333333333333, "true": 0.06666666666666667}, "group": {"0": 0.4, "1": 0.35555555555555557, "2": 0.24444444444444444}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"2/15\"],[\"blue\",\"7/45\"],[\"brown\",\"1/45\"],[\"cyan\",\"2/45\"],[\"gray\",\"1/45\"],[\"green\",\"1/15\"],[\"orange\",\"8/45\"],[\"pink\",\"8/45\"],[\"purple\",\"1/9\"],[\"red\",\"2/45\"],[\"white\",\"1/45\"],[\"yellow\",\"1/45\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0.01, "orange": 0.11, "blue": 0.03, "yellow": 0.01, "brown": 0, "black": 0.01, "gray": 0, "pink": 0.81, "cyan": 0, "purple": 0.01, "green": 0.01, "red": 0}, "focus": {"false": 0.9, "true": 0.1}, "group": {"0": 0.89, "1": 0.1, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:50", "state_id": "known_chance_v2:train:50", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: cyan=6, pink=6, purple=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"pink": "The sampled ball is pink", "purple": "The sampled ball is purple", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a purple ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: cyan", "The drawn color is one of: pink", "The drawn color is one of: purple"]}}, "gold": {}, "gold_probs": {"outcome": {"pink": 0.3157894736842105, "purple": 0.3684210526315789, "cyan": 0.3157894736842105}, "focus": {"false": 0.631578947368421, "true": 0.3684210526315789}, "group": {"0": 0.3157894736842105, "1": 0.3157894736842105, "2": 0.3684210526315789}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"6/19\"],[\"pink\",\"6/19\"],[\"purple\",\"7/19\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.88, "pink": 0.02, "cyan": 0.1}, "focus": {"false": 0.64, "true": 0.36}, "group": {"0": 0.05, "1": 0.02, "2": 0.93}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:47", "state_id": "known_chance_v2:train:47", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: yellow=2, red=8, gray=1, blue=4, pink=9. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "yellow": "The sampled ball is yellow", "red": "The sampled ball is red", "pink": "The sampled ball is pink", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: yellow, blue", "The drawn color is one of: red, pink", "The drawn color is one of: gray"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.16666666666666666, "yellow": 0.08333333333333333, "red": 0.3333333333333333, "pink": 0.375, "gray": 0.041666666666666664}, "focus": {"false": 0.9166666666666666, "true": 0.08333333333333333}, "group": {"0": 0.25, "1": 0.7083333333333333, "2": 0.041666666666666664}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"1/6\"],[\"gray\",\"1/24\"],[\"pink\",\"3/8\"],[\"red\",\"1/3\"],[\"yellow\",\"1/12\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"yellow": 0.02, "blue": 0, "red": 0.11, "pink": 0.86, "gray": 0.01}, "focus": {"false": 0.88, "true": 0.12}, "group": {"0": 0, "1": 1, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:52", "state_id": "known_chance_v2:train:52", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: white=6, brown=3, green=7, cyan=3, blue=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"white": "The sampled ball is white", "brown": "The sampled ball is brown", "green": "The sampled ball is green", "blue": "The sampled ball is blue", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a white ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white, cyan", "The drawn color is one of: brown, blue", "The drawn color is one of: green"]}}, "gold": {}, "gold_probs": {"outcome": {"white": 0.2222222222222222, "brown": 0.1111111111111111, "green": 0.25925925925925924, "blue": 0.2962962962962963, "cyan": 0.1111111111111111}, "focus": {"false": 0.7777777777777778, "true": 0.2222222222222222}, "group": {"0": 0.3333333333333333, "1": 0.4074074074074074, "2": 0.25925925925925924}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"8/27\"],[\"brown\",\"1/9\"],[\"cyan\",\"1/9\"],[\"green\",\"7/27\"],[\"white\",\"2/9\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0.08, "cyan": 0, "brown": 0, "blue": 0.89, "green": 0.03}, "focus": {"false": 0.81, "true": 0.19}, "group": {"0": 0.05, "1": 0.58, "2": 0.37}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:51", "state_id": "known_chance_v2:train:51", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: pink=5, gray=8, green=2, orange=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "orange": "The sampled ball is orange", "pink": "The sampled ball is pink", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a green ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: pink, orange", "The drawn color is one of: gray", "The drawn color is one of: green"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.1, "orange": 0.25, "pink": 0.25, "gray": 0.4}, "focus": {"false": 0.9, "true": 0.1}, "group": {"0": 0.5, "1": 0.4, "2": 0.1}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"gray\",\"2/5\"],[\"green\",\"1/10\"],[\"orange\",\"1/4\"],[\"pink\",\"1/4\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"gray": 0.91, "green": 0.01, "orange": 0.01, "pink": 0.07}, "focus": {"false": 0.88, "true": 0.12}, "group": {"0": 0.14, "1": 0.86, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:53", "state_id": "known_chance_v2:train:53", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: white=3, cyan=8, brown=8, pink=7, green=4, red=8, orange=2, black=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"brown": "The sampled ball is brown", "black": "The sampled ball is black", "orange": "The sampled ball is orange", "white": "The sampled ball is white", "green": "The sampled ball is green", "pink": "The sampled ball is pink", "cyan": "The sampled ball is cyan", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a red ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white, pink, orange", "The drawn color is one of: cyan, green, black", "The drawn color is one of: brown, red"]}}, "gold": {}, "gold_probs": {"outcome": {"brown": 0.19047619047619047, "black": 0.047619047619047616, "orange": 0.047619047619047616, "white": 0.07142857142857142, "green": 0.09523809523809523, "pink": 0.16666666666666666, "cyan": 0.19047619047619047, "red": 0.19047619047619047}, "focus": {"false": 0.8095238095238095, "true": 0.19047619047619047}, "group": {"0": 0.2857142857142857, "1": 0.3333333333333333, "2": 0.38095238095238093}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/21\"],[\"brown\",\"4/21\"],[\"cyan\",\"4/21\"],[\"green\",\"2/21\"],[\"orange\",\"1/21\"],[\"pink\",\"1/6\"],[\"red\",\"4/21\"],[\"white\",\"1/14\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0.13999999999999999, "orange": 0.02, "brown": 0.12, "black": 0.01, "pink": 0.04, "cyan": 0.53, "green": 0.01, "red": 0.13}, "focus": {"false": 0.81, "true": 0.19}, "group": {"0": 0.02, "1": 0.2, "2": 0.78}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:54", "state_id": "known_chance_v2:train:54", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: white=7, gray=8, blue=2, brown=8, black=7, purple=4, red=9, green=5, yellow=9, pink=9, orange=1, cyan=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"orange": "The sampled ball is orange", "yellow": "The sampled ball is yellow", "cyan": "The sampled ball is cyan", "pink": "The sampled ball is pink", "green": "The sampled ball is green", "black": "The sampled ball is black", "brown": "The sampled ball is brown", "gray": "The sampled ball is gray", "white": "The sampled ball is white", "blue": "The sampled ball is blue", "purple": "The sampled ball is purple", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white, brown, red, pink", "The drawn color is one of: gray, black, green, orange", "The drawn color is one of: blue, purple, yellow, cyan"]}}, "gold": {}, "gold_probs": {"outcome": {"orange": 0.012987012987012988, "yellow": 0.11688311688311688, "cyan": 0.1038961038961039, "pink": 0.11688311688311688, "green": 0.06493506493506493, "black": 0.09090909090909091, "brown": 0.1038961038961039, "gray": 0.1038961038961039, "white": 0.09090909090909091, "blue": 0.025974025974025976, "purple": 0.05194805194805195, "red": 0.11688311688311688}, "focus": {"false": 0.9090909090909091, "true": 0.09090909090909091}, "group": {"0": 0.4285714285714286, "1": 0.2727272727272727, "2": 0.2987012987012987}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/11\"],[\"blue\",\"2/77\"],[\"brown\",\"8/77\"],[\"cyan\",\"8/77\"],[\"gray\",\"8/77\"],[\"green\",\"5/77\"],[\"orange\",\"1/77\"],[\"pink\",\"9/77\"],[\"purple\",\"4/77\"],[\"red\",\"9/77\"],[\"white\",\"1/11\"],[\"yellow\",\"9/77\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"red": 0.29, "brown": 0.03, "blue": 0.02, "gray": 0.07, "black": 0.03, "yellow": 0.06, "orange": 0.1, "purple": 0.01, "white": 0.3, "cyan": 0.03, "green": 0.02, "pink": 0.04}, "focus": {"false": 0.84, "true": 0.16}, "group": {"0": 0.91, "1": 0.04, "2": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:55", "state_id": "known_chance_v2:train:55", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: cyan=1, green=3, blue=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "green": "The sampled ball is green", "blue": "The sampled ball is blue"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a green ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: cyan", "The drawn color is one of: green", "The drawn color is one of: blue"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.14285714285714285, "green": 0.42857142857142855, "blue": 0.42857142857142855}, "focus": {"false": 0.5714285714285714, "true": 0.42857142857142855}, "group": {"0": 0.14285714285714285, "1": 0.42857142857142855, "2": 0.42857142857142855}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"3/7\"],[\"cyan\",\"1/7\"],[\"green\",\"3/7\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0.38, "blue": 0.07, "green": 0.55}, "focus": {"false": 0.61, "true": 0.39}, "group": {"0": 0.02, "1": 0.89, "2": 0.09}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:56", "state_id": "known_chance_v2:train:56", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: cyan=2, red=8, blue=9, yellow=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "cyan": "The sampled ball is cyan", "red": "The sampled ball is red", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: cyan, yellow", "The drawn color is one of: red", "The drawn color is one of: blue"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.4090909090909091, "cyan": 0.09090909090909091, "red": 0.36363636363636365, "yellow": 0.13636363636363635}, "focus": {"false": 0.8636363636363636, "true": 0.13636363636363635}, "group": {"0": 0.22727272727272727, "1": 0.36363636363636365, "2": 0.4090909090909091}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"9/22\"],[\"cyan\",\"1/11\"],[\"red\",\"4/11\"],[\"yellow\",\"3/22\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0.04, "red": 0.1, "blue": 0.85, "yellow": 0.01}, "focus": {"false": 0.87, "true": 0.13}, "group": {"0": 0.01, "1": 0.05, "2": 0.93}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:58", "state_id": "known_chance_v2:train:58", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: yellow=3, orange=7, red=6, brown=4, blue=1, green=9, white=2, pink=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"orange": "The sampled ball is orange", "red": "The sampled ball is red", "yellow": "The sampled ball is yellow", "pink": "The sampled ball is pink", "white": "The sampled ball is white", "blue": "The sampled ball is blue", "green": "The sampled ball is green", "brown": "The sampled ball is brown"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: yellow, brown, white", "The drawn color is one of: orange, blue, pink", "The drawn color is one of: red, green"]}}, "gold": {}, "gold_probs": {"outcome": {"orange": 0.18421052631578946, "red": 0.15789473684210525, "yellow": 0.07894736842105263, "pink": 0.15789473684210525, "white": 0.05263157894736842, "blue": 0.02631578947368421, "green": 0.23684210526315788, "brown": 0.10526315789473684}, "focus": {"false": 0.9736842105263158, "true": 0.02631578947368421}, "group": {"0": 0.23684210526315788, "1": 0.3684210526315789, "2": 0.39473684210526316}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"1/38\"],[\"brown\",\"2/19\"],[\"green\",\"9/38\"],[\"orange\",\"7/38\"],[\"pink\",\"3/19\"],[\"red\",\"3/19\"],[\"white\",\"1/19\"],[\"yellow\",\"3/38\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0.95, "blue": 0.01, "pink": 0, "white": 0, "red": 0, "brown": 0, "orange": 0.02, "yellow": 0.02}, "focus": {"false": 0.9299999999999999, "true": 0.07}, "group": {"0": 0.01, "1": 0.05, "2": 0.94}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:57", "state_id": "known_chance_v2:train:57", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: gray=5, white=7, pink=6, brown=9, blue=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"pink": "The sampled ball is pink", "blue": "The sampled ball is blue", "brown": "The sampled ball is brown", "white": "The sampled ball is white", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: gray, brown", "The drawn color is one of: white, blue", "The drawn color is one of: pink"]}}, "gold": {}, "gold_probs": {"outcome": {"pink": 0.20689655172413793, "blue": 0.06896551724137931, "brown": 0.3103448275862069, "white": 0.2413793103448276, "gray": 0.1724137931034483}, "focus": {"false": 0.9310344827586207, "true": 0.06896551724137931}, "group": {"0": 0.4827586206896552, "1": 0.3103448275862069, "2": 0.20689655172413793}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"2/29\"],[\"brown\",\"9/29\"],[\"gray\",\"5/29\"],[\"pink\",\"6/29\"],[\"white\",\"7/29\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0.04, "gray": 0.03, "pink": 0, "blue": 0.01, "brown": 0.92}, "focus": {"false": 0.89, "true": 0.11}, "group": {"0": 0.93, "1": 0.05, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:60", "state_id": "known_chance_v2:train:60", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: yellow=2, blue=6, green=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"yellow": "The sampled ball is yellow", "blue": "The sampled ball is blue", "green": "The sampled ball is green"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: yellow", "The drawn color is one of: blue", "The drawn color is one of: green"]}}, "gold": {}, "gold_probs": {"outcome": {"yellow": 0.2, "blue": 0.6, "green": 0.2}, "focus": {"false": 0.4, "true": 0.6}, "group": {"0": 0.2, "1": 0.6, "2": 0.2}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"3/5\"],[\"green\",\"1/5\"],[\"yellow\",\"1/5\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0, "blue": 0.97, "yellow": 0.03}, "focus": {"false": 0.48, "true": 0.52}, "group": {"0": 0.01, "1": 0.99, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:59", "state_id": "known_chance_v2:train:59", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: brown=9, pink=9, blue=4, black=3, cyan=8, white=8, green=3, gray=2, orange=1, purple=2, yellow=7, red=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"yellow": "The sampled ball is yellow", "blue": "The sampled ball is blue", "brown": "The sampled ball is brown", "black": "The sampled ball is black", "white": "The sampled ball is white", "orange": "The sampled ball is orange", "cyan": "The sampled ball is cyan", "pink": "The sampled ball is pink", "gray": "The sampled ball is gray", "red": "The sampled ball is red", "purple": "The sampled ball is purple", "green": "The sampled ball is green"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a white ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: brown, black, green, purple", "The drawn color is one of: pink, cyan, gray, yellow", "The drawn color is one of: blue, white, orange, red"]}}, "gold": {}, "gold_probs": {"outcome": {"yellow": 0.109375, "blue": 0.0625, "brown": 0.140625, "black": 0.046875, "white": 0.125, "orange": 0.015625, "cyan": 0.125, "pink": 0.140625, "gray": 0.03125, "red": 0.125, "purple": 0.03125, "green": 0.046875}, "focus": {"false": 0.875, "true": 0.125}, "group": {"0": 0.265625, "1": 0.40625, "2": 0.328125}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"3/64\"],[\"blue\",\"1/16\"],[\"brown\",\"9/64\"],[\"cyan\",\"1/8\"],[\"gray\",\"1/32\"],[\"green\",\"3/64\"],[\"orange\",\"1/64\"],[\"pink\",\"9/64\"],[\"purple\",\"1/32\"],[\"red\",\"1/8\"],[\"white\",\"1/8\"],[\"yellow\",\"7/64\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.01, "pink": 0.15, "orange": 0.11, "green": 0.01, "blue": 0.01, "red": 0.02, "cyan": 0.01, "white": 0.01, "yellow": 0.01, "brown": 0.64, "black": 0.01, "gray": 0.01}, "focus": {"false": 0.86, "true": 0.14}, "group": {"0": 0.26, "1": 0.6, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:65", "state_id": "known_chance_v2:train:65", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: pink=3, yellow=8, gray=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"yellow": "The sampled ball is yellow", "pink": "The sampled ball is pink", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a pink ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: pink", "The drawn color is one of: yellow", "The drawn color is one of: gray"]}}, "gold": {}, "gold_probs": {"outcome": {"yellow": 0.47058823529411764, "pink": 0.17647058823529413, "gray": 0.35294117647058826}, "focus": {"false": 0.8235294117647058, "true": 0.17647058823529413}, "group": {"0": 0.17647058823529413, "1": 0.47058823529411764, "2": 0.35294117647058826}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"gray\",\"6/17\"],[\"pink\",\"3/17\"],[\"yellow\",\"8/17\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.04, "gray": 0.01, "yellow": 0.95}, "focus": {"false": 0.83, "true": 0.17}, "group": {"0": 0.01, "1": 0.97, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:64", "state_id": "known_chance_v2:train:64", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: white=5, red=1, brown=5, cyan=2, orange=4, gray=6, pink=8, green=8, black=4, blue=9, purple=4, yellow=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "red": "The sampled ball is red", "cyan": "The sampled ball is cyan", "white": "The sampled ball is white", "brown": "The sampled ball is brown", "blue": "The sampled ball is blue", "purple": "The sampled ball is purple", "pink": "The sampled ball is pink", "black": "The sampled ball is black", "orange": "The sampled ball is orange", "yellow": "The sampled ball is yellow", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white, cyan, pink, blue", "The drawn color is one of: red, orange, green, purple", "The drawn color is one of: brown, gray, black, yellow"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.13333333333333333, "red": 0.016666666666666666, "cyan": 0.03333333333333333, "white": 0.08333333333333333, "brown": 0.08333333333333333, "blue": 0.15, "purple": 0.06666666666666667, "pink": 0.13333333333333333, "black": 0.06666666666666667, "orange": 0.06666666666666667, "yellow": 0.06666666666666667, "gray": 0.1}, "focus": {"false": 0.9333333333333333, "true": 0.06666666666666667}, "group": {"0": 0.39999999999999997, "1": 0.2833333333333333, "2": 0.31666666666666665}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/15\"],[\"blue\",\"3/20\"],[\"brown\",\"1/12\"],[\"cyan\",\"1/30\"],[\"gray\",\"1/10\"],[\"green\",\"2/15\"],[\"orange\",\"1/15\"],[\"pink\",\"2/15\"],[\"purple\",\"1/15\"],[\"red\",\"1/60\"],[\"white\",\"1/12\"],[\"yellow\",\"1/15\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.06999999999999999, "white": 0.09, "cyan": 0, "yellow": 0, "green": 0.02, "brown": 0, "orange": 0, "blue": 0.76, "purple": 0, "red": 0.05, "gray": 0.01, "black": 0}, "focus": {"false": 0.87, "true": 0.13}, "group": {"0": 0.83, "1": 0.03, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:63", "state_id": "known_chance_v2:train:63", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: black=4, cyan=3, brown=6, white=5, blue=4, pink=5, yellow=1, gray=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"brown": "The sampled ball is brown", "blue": "The sampled ball is blue", "cyan": "The sampled ball is cyan", "yellow": "The sampled ball is yellow", "pink": "The sampled ball is pink", "black": "The sampled ball is black", "white": "The sampled ball is white", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: black, white, yellow", "The drawn color is one of: cyan, blue, gray", "The drawn color is one of: brown, pink"]}}, "gold": {}, "gold_probs": {"outcome": {"brown": 0.2, "blue": 0.13333333333333333, "cyan": 0.1, "yellow": 0.03333333333333333, "pink": 0.16666666666666666, "black": 0.13333333333333333, "white": 0.16666666666666666, "gray": 0.06666666666666667}, "focus": {"false": 0.8666666666666667, "true": 0.13333333333333333}, "group": {"0": 0.3333333333333333, "1": 0.3, "2": 0.3666666666666667}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"2/15\"],[\"blue\",\"2/15\"],[\"brown\",\"1/5\"],[\"cyan\",\"1/10\"],[\"gray\",\"1/15\"],[\"pink\",\"1/6\"],[\"white\",\"1/6\"],[\"yellow\",\"1/30\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"black": 0.03, "white": 0.01, "cyan": 0, "pink": 0.01, "gray": 0, "brown": 0.9, "yellow": 0.05, "blue": 0}, "focus": {"false": 0.84, "true": 0.16}, "group": {"0": 0.03, "1": 0.02, "2": 0.95}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:61", "state_id": "known_chance_v2:train:61", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: blue=2, red=8, brown=3, black=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"brown": "The sampled ball is brown", "black": "The sampled ball is black", "blue": "The sampled ball is blue", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue, black", "The drawn color is one of: red", "The drawn color is one of: brown"]}}, "gold": {}, "gold_probs": {"outcome": {"brown": 0.15789473684210525, "black": 0.3157894736842105, "blue": 0.10526315789473684, "red": 0.42105263157894735}, "focus": {"false": 0.6842105263157895, "true": 0.3157894736842105}, "group": {"0": 0.42105263157894735, "1": 0.42105263157894735, "2": 0.15789473684210525}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"6/19\"],[\"blue\",\"2/19\"],[\"brown\",\"3/19\"],[\"red\",\"8/19\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"red": 0.96, "blue": 0.03, "brown": 0, "black": 0.01}, "focus": {"false": 0.81, "true": 0.19}, "group": {"0": 0.02, "1": 0.97, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:62", "state_id": "known_chance_v2:train:62", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: red=5, purple=1, brown=9, pink=9, blue=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "purple": "The sampled ball is purple", "red": "The sampled ball is red", "pink": "The sampled ball is pink", "brown": "The sampled ball is brown"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a brown ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: red, pink", "The drawn color is one of: purple, blue", "The drawn color is one of: brown"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.1724137931034483, "purple": 0.034482758620689655, "red": 0.1724137931034483, "pink": 0.3103448275862069, "brown": 0.3103448275862069}, "focus": {"false": 0.6896551724137931, "true": 0.3103448275862069}, "group": {"0": 0.4827586206896552, "1": 0.20689655172413796, "2": 0.3103448275862069}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"5/29\"],[\"brown\",\"9/29\"],[\"pink\",\"9/29\"],[\"purple\",\"1/29\"],[\"red\",\"5/29\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"blue": 0.01, "pink": 0.13, "brown": 0.77, "purple": 0.04, "red": 0.05}, "focus": {"false": 0.73, "true": 0.27}, "group": {"0": 0.45, "1": 0.02, "2": 0.53}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:66", "state_id": "known_chance_v2:train:66", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: gray=6, yellow=1, white=4, red=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"yellow": "The sampled ball is yellow", "red": "The sampled ball is red", "white": "The sampled ball is white", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: gray, red", "The drawn color is one of: yellow", "The drawn color is one of: white"]}}, "gold": {}, "gold_probs": {"outcome": {"yellow": 0.08333333333333333, "red": 0.08333333333333333, "white": 0.3333333333333333, "gray": 0.5}, "focus": {"false": 0.9166666666666666, "true": 0.08333333333333333}, "group": {"0": 0.5833333333333334, "1": 0.08333333333333333, "2": 0.3333333333333333}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"gray\",\"1/2\"],[\"red\",\"1/12\"],[\"white\",\"1/3\"],[\"yellow\",\"1/12\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"red": 0, "gray": 0.98, "yellow": 0, "white": 0.02}, "focus": {"false": 0.89, "true": 0.11}, "group": {"0": 0.9, "1": 0.01, "2": 0.09}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:67", "state_id": "known_chance_v2:train:67", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: blue=1, white=9, purple=2, red=1, orange=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"orange": "The sampled ball is orange", "white": "The sampled ball is white", "blue": "The sampled ball is blue", "purple": "The sampled ball is purple", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue, red", "The drawn color is one of: white, orange", "The drawn color is one of: purple"]}}, "gold": {}, "gold_probs": {"outcome": {"orange": 0.07142857142857142, "white": 0.6428571428571429, "blue": 0.07142857142857142, "purple": 0.14285714285714285, "red": 0.07142857142857142}, "focus": {"false": 0.9285714285714286, "true": 0.07142857142857142}, "group": {"0": 0.14285714285714285, "1": 0.7142857142857143, "2": 0.14285714285714285}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"1/14\"],[\"orange\",\"1/14\"],[\"purple\",\"1/7\"],[\"red\",\"1/14\"],[\"white\",\"9/14\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"red": 0, "purple": 0.01, "white": 0.98, "blue": 0.01, "orange": 0}, "focus": {"false": 0.92, "true": 0.08}, "group": {"0": 0.02, "1": 0.95, "2": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:69", "state_id": "known_chance_v2:train:69", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: black=4, pink=2, purple=4, red=9, brown=7, blue=6, green=8, gray=1, yellow=9, orange=7, cyan=9, white=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "blue": "The sampled ball is blue", "orange": "The sampled ball is orange", "white": "The sampled ball is white", "red": "The sampled ball is red", "purple": "The sampled ball is purple", "black": "The sampled ball is black", "green": "The sampled ball is green", "brown": "The sampled ball is brown", "pink": "The sampled ball is pink", "gray": "The sampled ball is gray", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: black, red, green, orange", "The drawn color is one of: pink, brown, gray, cyan", "The drawn color is one of: purple, blue, yellow, white"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.13432835820895522, "blue": 0.08955223880597014, "orange": 0.1044776119402985, "white": 0.014925373134328358, "red": 0.13432835820895522, "purple": 0.05970149253731343, "black": 0.05970149253731343, "green": 0.11940298507462686, "brown": 0.1044776119402985, "pink": 0.029850746268656716, "gray": 0.014925373134328358, "yellow": 0.13432835820895522}, "focus": {"false": 0.8656716417910448, "true": 0.13432835820895522}, "group": {"0": 0.417910447761194, "1": 0.2835820895522388, "2": 0.29850746268656714}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"4/67\"],[\"blue\",\"6/67\"],[\"brown\",\"7/67\"],[\"cyan\",\"9/67\"],[\"gray\",\"1/67\"],[\"green\",\"8/67\"],[\"orange\",\"7/67\"],[\"pink\",\"2/67\"],[\"purple\",\"4/67\"],[\"red\",\"9/67\"],[\"white\",\"1/67\"],[\"yellow\",\"9/67\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0.01, "pink": 0.01, "yellow": 0.07, "red": 0.76, "cyan": 0.03, "green": 0.02, "black": 0.05, "brown": 0, "gray": 0.03, "purple": 0.01, "orange": 0, "blue": 0.01}, "focus": {"false": 0.83, "true": 0.17}, "group": {"0": 0.82, "1": 0.03, "2": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:68", "state_id": "known_chance_v2:train:68", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: cyan=8, brown=6, black=3, purple=5, blue=1, gray=3, green=6, red=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "brown": "The sampled ball is brown", "red": "The sampled ball is red", "cyan": "The sampled ball is cyan", "green": "The sampled ball is green", "purple": "The sampled ball is purple", "gray": "The sampled ball is gray", "blue": "The sampled ball is blue"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: cyan, purple, green", "The drawn color is one of: brown, blue, red", "The drawn color is one of: black, gray"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.08823529411764706, "brown": 0.17647058823529413, "red": 0.058823529411764705, "cyan": 0.23529411764705882, "green": 0.17647058823529413, "purple": 0.14705882352941177, "gray": 0.08823529411764706, "blue": 0.029411764705882353}, "focus": {"false": 0.7647058823529411, "true": 0.23529411764705882}, "group": {"0": 0.5588235294117647, "1": 0.2647058823529412, "2": 0.17647058823529413}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"3/34\"],[\"blue\",\"1/34\"],[\"brown\",\"3/17\"],[\"cyan\",\"4/17\"],[\"gray\",\"3/34\"],[\"green\",\"3/17\"],[\"purple\",\"5/34\"],[\"red\",\"1/17\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0, "cyan": 0.96, "brown": 0.01, "blue": 0.03, "gray": 0, "red": 0, "green": 0, "black": 0}, "focus": {"false": 0.81, "true": 0.19}, "group": {"0": 0.99, "1": 0, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:71", "state_id": "known_chance_v2:train:71", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: red=6, cyan=5, orange=8, black=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "orange": "The sampled ball is orange", "red": "The sampled ball is red", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: red, black", "The drawn color is one of: cyan", "The drawn color is one of: orange"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.13636363636363635, "orange": 0.36363636363636365, "red": 0.2727272727272727, "cyan": 0.22727272727272727}, "focus": {"false": 0.7727272727272727, "true": 0.22727272727272727}, "group": {"0": 0.40909090909090906, "1": 0.22727272727272727, "2": 0.36363636363636365}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"3/22\"],[\"cyan\",\"5/22\"],[\"orange\",\"4/11\"],[\"red\",\"3/11\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0.01, "black": 0.01, "orange": 0.84, "red": 0.14}, "focus": {"false": 0.8200000000000001, "true": 0.18}, "group": {"0": 0.03, "1": 0.01, "2": 0.96}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:70", "state_id": "known_chance_v2:train:70", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: orange=7, blue=6, yellow=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "yellow": "The sampled ball is yellow", "orange": "The sampled ball is orange"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: orange", "The drawn color is one of: blue", "The drawn color is one of: yellow"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.3333333333333333, "yellow": 0.2777777777777778, "orange": 0.3888888888888889}, "focus": {"false": 0.7222222222222222, "true": 0.2777777777777778}, "group": {"0": 0.3888888888888889, "1": 0.3333333333333333, "2": 0.2777777777777778}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"1/3\"],[\"orange\",\"7/18\"],[\"yellow\",\"5/18\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"blue": 0.01, "orange": 0.98, "yellow": 0.01}, "focus": {"false": 0.79, "true": 0.21}, "group": {"0": 0.96, "1": 0.02, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:73", "state_id": "known_chance_v2:train:73", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: cyan=6, yellow=9, brown=4, blue=3, orange=6, green=4, pink=3, purple=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"purple": "The sampled ball is purple", "pink": "The sampled ball is pink", "blue": "The sampled ball is blue", "green": "The sampled ball is green", "yellow": "The sampled ball is yellow", "brown": "The sampled ball is brown", "orange": "The sampled ball is orange", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a brown ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: cyan, blue, pink", "The drawn color is one of: yellow, orange, purple", "The drawn color is one of: brown, green"]}}, "gold": {}, "gold_probs": {"outcome": {"purple": 0.16666666666666666, "pink": 0.07142857142857142, "blue": 0.07142857142857142, "green": 0.09523809523809523, "yellow": 0.21428571428571427, "brown": 0.09523809523809523, "orange": 0.14285714285714285, "cyan": 0.14285714285714285}, "focus": {"false": 0.9047619047619048, "true": 0.09523809523809523}, "group": {"0": 0.2857142857142857, "1": 0.5238095238095237, "2": 0.19047619047619047}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"1/14\"],[\"brown\",\"2/21\"],[\"cyan\",\"1/7\"],[\"green\",\"2/21\"],[\"orange\",\"1/7\"],[\"pink\",\"1/14\"],[\"purple\",\"1/6\"],[\"yellow\",\"3/14\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0, "purple": 0.01, "yellow": 0.96, "cyan": 0.03, "blue": 0, "orange": 0, "green": 0, "brown": 0}, "focus": {"false": 0.83, "true": 0.17}, "group": {"0": 0.01, "1": 0.98, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:75", "state_id": "known_chance_v2:train:75", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: cyan=6, green=5, orange=9. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "cyan": "The sampled ball is cyan", "orange": "The sampled ball is orange"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: cyan", "The drawn color is one of: green", "The drawn color is one of: orange"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.25, "cyan": 0.3, "orange": 0.45}, "focus": {"false": 0.7, "true": 0.3}, "group": {"0": 0.3, "1": 0.25, "2": 0.45}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"3/10\"],[\"green\",\"1/4\"],[\"orange\",\"9/20\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"orange": 0.73, "green": 0.03, "cyan": 0.24}, "focus": {"false": 0.75, "true": 0.25}, "group": {"0": 0.05, "1": 0.01, "2": 0.94}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:72", "state_id": "known_chance_v2:train:72", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: purple=7, gray=4, orange=6, blue=8, green=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "gray": "The sampled ball is gray", "purple": "The sampled ball is purple", "orange": "The sampled ball is orange", "green": "The sampled ball is green"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: purple, blue", "The drawn color is one of: gray, green", "The drawn color is one of: orange"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.25806451612903225, "gray": 0.12903225806451613, "purple": 0.22580645161290322, "orange": 0.1935483870967742, "green": 0.1935483870967742}, "focus": {"false": 0.7419354838709677, "true": 0.25806451612903225}, "group": {"0": 0.4838709677419355, "1": 0.3225806451612903, "2": 0.1935483870967742}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"8/31\"],[\"gray\",\"4/31\"],[\"green\",\"6/31\"],[\"orange\",\"6/31\"],[\"purple\",\"7/31\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0.01, "orange": 0.01, "blue": 0.79, "gray": 0.02, "purple": 0.17}, "focus": {"false": 0.78, "true": 0.22}, "group": {"0": 0.99, "1": 0.01, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:74", "state_id": "known_chance_v2:train:74", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: yellow=1, brown=1, pink=1, orange=3, blue=7, cyan=3, black=3, gray=4, red=3, white=3, green=7, purple=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"white": "The sampled ball is white", "brown": "The sampled ball is brown", "pink": "The sampled ball is pink", "cyan": "The sampled ball is cyan", "purple": "The sampled ball is purple", "green": "The sampled ball is green", "yellow": "The sampled ball is yellow", "black": "The sampled ball is black", "blue": "The sampled ball is blue", "orange": "The sampled ball is orange", "gray": "The sampled ball is gray", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a brown ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: yellow, orange, black, white", "The drawn color is one of: brown, blue, gray, green", "The drawn color is one of: pink, cyan, red, purple"]}}, "gold": {}, "gold_probs": {"outcome": {"white": 0.08108108108108109, "brown": 0.02702702702702703, "pink": 0.02702702702702703, "cyan": 0.08108108108108109, "purple": 0.02702702702702703, "green": 0.1891891891891892, "yellow": 0.02702702702702703, "black": 0.08108108108108109, "blue": 0.1891891891891892, "orange": 0.08108108108108109, "gray": 0.10810810810810811, "red": 0.08108108108108109}, "focus": {"false": 0.972972972972973, "true": 0.02702702702702703}, "group": {"0": 0.2702702702702703, "1": 0.5135135135135136, "2": 0.21621621621621623}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"3/37\"],[\"blue\",\"7/37\"],[\"brown\",\"1/37\"],[\"cyan\",\"3/37\"],[\"gray\",\"4/37\"],[\"green\",\"7/37\"],[\"orange\",\"3/37\"],[\"pink\",\"1/37\"],[\"purple\",\"1/37\"],[\"red\",\"3/37\"],[\"white\",\"3/37\"],[\"yellow\",\"1/37\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0.16, "pink": 0, "red": 0, "blue": 0.73, "orange": 0.01, "yellow": 0.03, "brown": 0.01, "gray": 0.04, "black": 0, "cyan": 0.01, "white": 0.01, "purple": 0}, "focus": {"false": 0.9299999999999999, "true": 0.07}, "group": {"0": 0.01, "1": 0.98, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:76", "state_id": "known_chance_v2:train:76", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: gray=3, pink=1, cyan=9, green=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"gray": "The sampled ball is gray", "cyan": "The sampled ball is cyan", "green": "The sampled ball is green", "pink": "The sampled ball is pink"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a green ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: gray, green", "The drawn color is one of: pink", "The drawn color is one of: cyan"]}}, "gold": {}, "gold_probs": {"outcome": {"gray": 0.15, "cyan": 0.45, "green": 0.35, "pink": 0.05}, "focus": {"false": 0.65, "true": 0.35}, "group": {"0": 0.5, "1": 0.05, "2": 0.45}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"9/20\"],[\"gray\",\"3/20\"],[\"green\",\"7/20\"],[\"pink\",\"1/20\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"gray": 0.02, "pink": 0.01, "cyan": 0.96, "green": 0.01}, "focus": {"false": 0.78, "true": 0.22}, "group": {"0": 0.02, "1": 0, "2": 0.98}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:77", "state_id": "known_chance_v2:train:77", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: gray=1, white=9, yellow=6, cyan=9, black=9. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "black": "The sampled ball is black", "white": "The sampled ball is white", "gray": "The sampled ball is gray", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: gray, cyan", "The drawn color is one of: white, black", "The drawn color is one of: yellow"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.2647058823529412, "black": 0.2647058823529412, "white": 0.2647058823529412, "gray": 0.029411764705882353, "yellow": 0.17647058823529413}, "focus": {"false": 0.8235294117647058, "true": 0.17647058823529413}, "group": {"0": 0.29411764705882354, "1": 0.5294117647058824, "2": 0.17647058823529413}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"9/34\"],[\"cyan\",\"9/34\"],[\"gray\",\"1/34\"],[\"white\",\"9/34\"],[\"yellow\",\"3/17\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0.72, "cyan": 0.08, "yellow": 0.02, "gray": 0.14, "black": 0.04}, "focus": {"false": 0.81, "true": 0.19}, "group": {"0": 0.16, "1": 0.82, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:79", "state_id": "known_chance_v2:train:79", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: purple=6, yellow=7, cyan=1, green=8, red=6, black=8, white=7, gray=8, orange=1, brown=7, pink=1, blue=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "brown": "The sampled ball is brown", "yellow": "The sampled ball is yellow", "gray": "The sampled ball is gray", "blue": "The sampled ball is blue", "pink": "The sampled ball is pink", "black": "The sampled ball is black", "orange": "The sampled ball is orange", "purple": "The sampled ball is purple", "white": "The sampled ball is white", "green": "The sampled ball is green", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a brown ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: purple, green, white, brown", "The drawn color is one of: yellow, red, gray, pink", "The drawn color is one of: cyan, black, orange, blue"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.016129032258064516, "brown": 0.11290322580645161, "yellow": 0.11290322580645161, "gray": 0.12903225806451613, "blue": 0.03225806451612903, "pink": 0.016129032258064516, "black": 0.12903225806451613, "orange": 0.016129032258064516, "purple": 0.0967741935483871, "white": 0.11290322580645161, "green": 0.12903225806451613, "red": 0.0967741935483871}, "focus": {"false": 0.8870967741935484, "true": 0.11290322580645161}, "group": {"0": 0.45161290322580644, "1": 0.3548387096774194, "2": 0.1935483870967742}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"4/31\"],[\"blue\",\"1/31\"],[\"brown\",\"7/62\"],[\"cyan\",\"1/62\"],[\"gray\",\"4/31\"],[\"green\",\"4/31\"],[\"orange\",\"1/62\"],[\"pink\",\"1/62\"],[\"purple\",\"3/31\"],[\"red\",\"3/31\"],[\"white\",\"7/62\"],[\"yellow\",\"7/62\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0.49, "orange": 0.02, "white": 0.01, "purple": 0.13, "gray": 0.09, "brown": 0.01, "yellow": 0.03, "pink": 0.01, "cyan": 0.02, "black": 0.16, "blue": 0.01, "red": 0.02}, "focus": {"false": 0.85, "true": 0.15}, "group": {"0": 0.77, "1": 0.09, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:78", "state_id": "known_chance_v2:train:78", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: red=3, cyan=8, black=7, white=8, orange=4, purple=6, blue=5, brown=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "red": "The sampled ball is red", "brown": "The sampled ball is brown", "purple": "The sampled ball is purple", "white": "The sampled ball is white", "orange": "The sampled ball is orange", "black": "The sampled ball is black", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a purple ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: red, white, blue", "The drawn color is one of: cyan, orange, brown", "The drawn color is one of: black, purple"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.10416666666666667, "red": 0.0625, "brown": 0.14583333333333334, "purple": 0.125, "white": 0.16666666666666666, "orange": 0.08333333333333333, "black": 0.14583333333333334, "cyan": 0.16666666666666666}, "focus": {"false": 0.875, "true": 0.125}, "group": {"0": 0.3333333333333333, "1": 0.3958333333333333, "2": 0.27083333333333337}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"7/48\"],[\"blue\",\"5/48\"],[\"brown\",\"7/48\"],[\"cyan\",\"1/6\"],[\"orange\",\"1/12\"],[\"purple\",\"1/8\"],[\"red\",\"1/16\"],[\"white\",\"1/6\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.01, "red": 0.18, "black": 0.02, "cyan": 0.61, "white": 0.13, "blue": 0.02, "brown": 0.02, "orange": 0.01}, "focus": {"false": 0.87, "true": 0.13}, "group": {"0": 0.12, "1": 0.78, "2": 0.1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:80", "state_id": "known_chance_v2:train:80", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: cyan=4, pink=2, red=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"red": "The sampled ball is red", "pink": "The sampled ball is pink", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: cyan", "The drawn color is one of: pink", "The drawn color is one of: red"]}}, "gold": {}, "gold_probs": {"outcome": {"red": 0.5384615384615384, "pink": 0.15384615384615385, "cyan": 0.3076923076923077}, "focus": {"false": 0.6923076923076923, "true": 0.3076923076923077}, "group": {"0": 0.3076923076923077, "1": 0.15384615384615385, "2": 0.5384615384615384}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"4/13\"],[\"pink\",\"2/13\"],[\"red\",\"7/13\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.01, "red": 0.94, "cyan": 0.05}, "focus": {"false": 0.77, "true": 0.23}, "group": {"0": 0.03, "1": 0.01, "2": 0.96}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:82", "state_id": "known_chance_v2:train:82", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: yellow=1, cyan=1, brown=5, green=3, white=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"brown": "The sampled ball is brown", "yellow": "The sampled ball is yellow", "green": "The sampled ball is green", "cyan": "The sampled ball is cyan", "white": "The sampled ball is white"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a brown ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: yellow, green", "The drawn color is one of: cyan, white", "The drawn color is one of: brown"]}}, "gold": {}, "gold_probs": {"outcome": {"brown": 0.4166666666666667, "yellow": 0.08333333333333333, "green": 0.25, "cyan": 0.08333333333333333, "white": 0.16666666666666666}, "focus": {"false": 0.5833333333333333, "true": 0.4166666666666667}, "group": {"0": 0.3333333333333333, "1": 0.25, "2": 0.4166666666666667}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"5/12\"],[\"cyan\",\"1/12\"],[\"green\",\"1/4\"],[\"white\",\"1/6\"],[\"yellow\",\"1/12\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"brown": 0.97, "green": 0.01, "white": 0, "cyan": 0, "yellow": 0.02}, "focus": {"false": 0.64, "true": 0.36}, "group": {"0": 0.01, "1": 0.01, "2": 0.98}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:81", "state_id": "known_chance_v2:train:81", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: gray=5, white=7, yellow=1, pink=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"white": "The sampled ball is white", "yellow": "The sampled ball is yellow", "pink": "The sampled ball is pink", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a white ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: gray, pink", "The drawn color is one of: white", "The drawn color is one of: yellow"]}}, "gold": {}, "gold_probs": {"outcome": {"white": 0.4375, "yellow": 0.0625, "pink": 0.1875, "gray": 0.3125}, "focus": {"false": 0.5625, "true": 0.4375}, "group": {"0": 0.5, "1": 0.4375, "2": 0.0625}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"gray\",\"5/16\"],[\"pink\",\"3/16\"],[\"white\",\"7/16\"],[\"yellow\",\"1/16\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"gray": 0.14, "pink": 0.01, "white": 0.83, "yellow": 0.02}, "focus": {"false": 0.69, "true": 0.31}, "group": {"0": 0.16, "1": 0.83, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:83", "state_id": "known_chance_v2:train:83", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: yellow=9, red=6, white=1, orange=5, black=8, pink=5, brown=2, gray=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"yellow": "The sampled ball is yellow", "white": "The sampled ball is white", "brown": "The sampled ball is brown", "pink": "The sampled ball is pink", "gray": "The sampled ball is gray", "red": "The sampled ball is red", "black": "The sampled ball is black", "orange": "The sampled ball is orange"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a orange ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: yellow, orange, brown", "The drawn color is one of: red, black, gray", "The drawn color is one of: white, pink"]}}, "gold": {}, "gold_probs": {"outcome": {"yellow": 0.23684210526315788, "white": 0.02631578947368421, "brown": 0.05263157894736842, "pink": 0.13157894736842105, "gray": 0.05263157894736842, "red": 0.15789473684210525, "black": 0.21052631578947367, "orange": 0.13157894736842105}, "focus": {"false": 0.868421052631579, "true": 0.13157894736842105}, "group": {"0": 0.42105263157894735, "1": 0.42105263157894735, "2": 0.15789473684210525}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"4/19\"],[\"brown\",\"1/19\"],[\"gray\",\"1/19\"],[\"orange\",\"5/38\"],[\"pink\",\"5/38\"],[\"red\",\"3/19\"],[\"white\",\"1/38\"],[\"yellow\",\"9/38\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0.02, "black": 0.02, "orange": 0, "yellow": 0.95, "brown": 0, "pink": 0, "red": 0.01, "gray": 0}, "focus": {"false": 0.85, "true": 0.15}, "group": {"0": 0.69, "1": 0.3, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:84", "state_id": "known_chance_v2:train:84", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: red=3, purple=5, blue=5, brown=4, yellow=4, gray=2, cyan=9, green=3, orange=1, pink=6, black=1, white=9. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"red": "The sampled ball is red", "cyan": "The sampled ball is cyan", "brown": "The sampled ball is brown", "blue": "The sampled ball is blue", "black": "The sampled ball is black", "purple": "The sampled ball is purple", "white": "The sampled ball is white", "green": "The sampled ball is green", "orange": "The sampled ball is orange", "gray": "The sampled ball is gray", "pink": "The sampled ball is pink", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a brown ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: red, brown, cyan, pink", "The drawn color is one of: purple, yellow, green, black", "The drawn color is one of: blue, gray, orange, white"]}}, "gold": {}, "gold_probs": {"outcome": {"red": 0.057692307692307696, "cyan": 0.17307692307692307, "brown": 0.07692307692307693, "blue": 0.09615384615384616, "black": 0.019230769230769232, "purple": 0.09615384615384616, "white": 0.17307692307692307, "green": 0.057692307692307696, "orange": 0.019230769230769232, "gray": 0.038461538461538464, "pink": 0.11538461538461539, "yellow": 0.07692307692307693}, "focus": {"false": 0.9230769230769231, "true": 0.07692307692307693}, "group": {"0": 0.4230769230769231, "1": 0.25, "2": 0.3269230769230769}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/52\"],[\"blue\",\"5/52\"],[\"brown\",\"1/13\"],[\"cyan\",\"9/52\"],[\"gray\",\"1/26\"],[\"green\",\"3/52\"],[\"orange\",\"1/52\"],[\"pink\",\"3/26\"],[\"purple\",\"5/52\"],[\"red\",\"3/52\"],[\"white\",\"9/52\"],[\"yellow\",\"1/13\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.01, "cyan": 0.58, "brown": 0.01, "pink": 0.03, "blue": 0.02, "gray": 0.01, "white": 0.23, "yellow": 0.01, "red": 0.06, "orange": 0.02, "green": 0.01, "black": 0.01}, "focus": {"false": 0.87, "true": 0.13}, "group": {"0": 0.67, "1": 0.03, "2": 0.3}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:86", "state_id": "known_chance_v2:train:86", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: blue=3, white=6, gray=5, pink=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "pink": "The sampled ball is pink", "white": "The sampled ball is white", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a pink ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue, pink", "The drawn color is one of: white", "The drawn color is one of: gray"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.15, "pink": 0.3, "white": 0.3, "gray": 0.25}, "focus": {"false": 0.7, "true": 0.3}, "group": {"0": 0.44999999999999996, "1": 0.3, "2": 0.25}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"3/20\"],[\"gray\",\"1/4\"],[\"pink\",\"3/10\"],[\"white\",\"3/10\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.08, "gray": 0.04, "blue": 0.32, "white": 0.56}, "focus": {"false": 0.78, "true": 0.22}, "group": {"0": 0.18, "1": 0.73, "2": 0.09}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:87", "state_id": "known_chance_v2:train:87", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: red=1, blue=8, brown=6, white=2, purple=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "red": "The sampled ball is red", "white": "The sampled ball is white", "purple": "The sampled ball is purple", "brown": "The sampled ball is brown"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: red, white", "The drawn color is one of: blue, purple", "The drawn color is one of: brown"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.42105263157894735, "red": 0.05263157894736842, "white": 0.10526315789473684, "purple": 0.10526315789473684, "brown": 0.3157894736842105}, "focus": {"false": 0.5789473684210527, "true": 0.42105263157894735}, "group": {"0": 0.15789473684210525, "1": 0.5263157894736842, "2": 0.3157894736842105}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"8/19\"],[\"brown\",\"6/19\"],[\"purple\",\"2/19\"],[\"red\",\"1/19\"],[\"white\",\"2/19\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"red": 0.03, "brown": 0.03, "purple": 0, "white": 0, "blue": 0.9400000000000001}, "focus": {"false": 0.7, "true": 0.3}, "group": {"0": 0.01, "1": 0.87, "2": 0.12}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:85", "state_id": "known_chance_v2:train:85", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: pink=5, orange=3, red=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"orange": "The sampled ball is orange", "pink": "The sampled ball is pink", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a orange ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: pink", "The drawn color is one of: orange", "The drawn color is one of: red"]}}, "gold": {}, "gold_probs": {"outcome": {"orange": 0.25, "pink": 0.4166666666666667, "red": 0.3333333333333333}, "focus": {"false": 0.75, "true": 0.25}, "group": {"0": 0.4166666666666667, "1": 0.25, "2": 0.3333333333333333}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"orange\",\"1/4\"],[\"pink\",\"5/12\"],[\"red\",\"1/3\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.98, "orange": 0.01, "red": 0.01}, "focus": {"false": 0.74, "true": 0.26}, "group": {"0": 0.96, "1": 0.01, "2": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:88", "state_id": "known_chance_v2:train:88", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: purple=4, brown=1, gray=2, black=6, pink=7, orange=1, white=1, red=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"purple": "The sampled ball is purple", "pink": "The sampled ball is pink", "black": "The sampled ball is black", "white": "The sampled ball is white", "brown": "The sampled ball is brown", "red": "The sampled ball is red", "gray": "The sampled ball is gray", "orange": "The sampled ball is orange"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a gray ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: purple, black, white", "The drawn color is one of: brown, pink, red", "The drawn color is one of: gray, orange"]}}, "gold": {}, "gold_probs": {"outcome": {"purple": 0.13333333333333333, "pink": 0.23333333333333334, "black": 0.2, "white": 0.03333333333333333, "brown": 0.03333333333333333, "red": 0.26666666666666666, "gray": 0.06666666666666667, "orange": 0.03333333333333333}, "focus": {"false": 0.9333333333333333, "true": 0.06666666666666667}, "group": {"0": 0.3666666666666667, "1": 0.5333333333333333, "2": 0.1}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/5\"],[\"brown\",\"1/30\"],[\"gray\",\"1/15\"],[\"orange\",\"1/30\"],[\"pink\",\"7/30\"],[\"purple\",\"2/15\"],[\"red\",\"4/15\"],[\"white\",\"1/30\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"brown": 0, "red": 0.97, "pink": 0.01, "orange": 0, "gray": 0, "black": 0.01, "white": 0, "purple": 0.01}, "focus": {"false": 0.87, "true": 0.13}, "group": {"0": 0.04, "1": 0.96, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:89", "state_id": "known_chance_v2:train:89", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: gray=7, purple=5, red=4, cyan=6, yellow=7, white=2, blue=2, brown=5, black=8, pink=6, orange=9, green=9. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "red": "The sampled ball is red", "yellow": "The sampled ball is yellow", "orange": "The sampled ball is orange", "brown": "The sampled ball is brown", "gray": "The sampled ball is gray", "black": "The sampled ball is black", "cyan": "The sampled ball is cyan", "green": "The sampled ball is green", "pink": "The sampled ball is pink", "white": "The sampled ball is white", "purple": "The sampled ball is purple"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: gray, cyan, blue, pink", "The drawn color is one of: purple, yellow, brown, orange", "The drawn color is one of: red, white, black, green"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.02857142857142857, "red": 0.05714285714285714, "yellow": 0.1, "orange": 0.12857142857142856, "brown": 0.07142857142857142, "gray": 0.1, "black": 0.11428571428571428, "cyan": 0.08571428571428572, "green": 0.12857142857142856, "pink": 0.08571428571428572, "white": 0.02857142857142857, "purple": 0.07142857142857142}, "focus": {"false": 0.9142857142857143, "true": 0.08571428571428572}, "group": {"0": 0.3, "1": 0.37142857142857144, "2": 0.32857142857142857}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"4/35\"],[\"blue\",\"1/35\"],[\"brown\",\"1/14\"],[\"cyan\",\"3/35\"],[\"gray\",\"1/10\"],[\"green\",\"9/70\"],[\"orange\",\"9/70\"],[\"pink\",\"3/35\"],[\"purple\",\"1/14\"],[\"red\",\"2/35\"],[\"white\",\"1/35\"],[\"yellow\",\"1/10\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"red": 0.02, "brown": 0.01, "blue": 0.02, "purple": 0.01, "black": 0.19999999999999998, "yellow": 0.02, "orange": 0.33999999999999997, "gray": 0.15, "cyan": 0.01, "white": 0.03, "pink": 0.01, "green": 0.18}, "focus": {"false": 0.88, "true": 0.12}, "group": {"0": 0.07, "1": 0.64, "2": 0.29}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:91", "state_id": "known_chance_v2:train:91", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: blue=1, cyan=8, gray=6, green=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "green": "The sampled ball is green", "gray": "The sampled ball is gray", "blue": "The sampled ball is blue"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue, green", "The drawn color is one of: cyan", "The drawn color is one of: gray"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.5, "green": 0.0625, "gray": 0.375, "blue": 0.0625}, "focus": {"false": 0.5, "true": 0.5}, "group": {"0": 0.125, "1": 0.5, "2": 0.375}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"1/16\"],[\"cyan\",\"1/2\"],[\"gray\",\"3/8\"],[\"green\",\"1/16\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"gray": 0.04, "green": 0, "blue": 0.01, "cyan": 0.95}, "focus": {"false": 0.65, "true": 0.35}, "group": {"0": 0.01, "1": 0.88, "2": 0.11}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:90", "state_id": "known_chance_v2:train:90", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: black=3, gray=1, blue=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "blue": "The sampled ball is blue", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: black", "The drawn color is one of: gray", "The drawn color is one of: blue"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.42857142857142855, "blue": 0.42857142857142855, "gray": 0.14285714285714285}, "focus": {"false": 0.5714285714285714, "true": 0.42857142857142855}, "group": {"0": 0.42857142857142855, "1": 0.14285714285714285, "2": 0.42857142857142855}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"3/7\"],[\"blue\",\"3/7\"],[\"gray\",\"1/7\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"gray": 0.04, "black": 0.93, "blue": 0.03}, "focus": {"false": 0.63, "true": 0.37}, "group": {"0": 0.87, "1": 0.04, "2": 0.09}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:92", "state_id": "known_chance_v2:train:92", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: blue=3, black=2, purple=6, red=2, orange=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"purple": "The sampled ball is purple", "blue": "The sampled ball is blue", "orange": "The sampled ball is orange", "red": "The sampled ball is red", "black": "The sampled ball is black"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a red ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue, red", "The drawn color is one of: black, orange", "The drawn color is one of: purple"]}}, "gold": {}, "gold_probs": {"outcome": {"purple": 0.42857142857142855, "blue": 0.21428571428571427, "orange": 0.07142857142857142, "red": 0.14285714285714285, "black": 0.14285714285714285}, "focus": {"false": 0.8571428571428572, "true": 0.14285714285714285}, "group": {"0": 0.3571428571428571, "1": 0.21428571428571427, "2": 0.42857142857142855}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/7\"],[\"blue\",\"3/14\"],[\"orange\",\"1/14\"],[\"purple\",\"3/7\"],[\"red\",\"1/7\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"orange": 0.01, "red": 0, "blue": 0.04, "purple": 0.95, "black": 0}, "focus": {"false": 0.85, "true": 0.15}, "group": {"0": 0.03, "1": 0.02, "2": 0.95}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:93", "state_id": "known_chance_v2:train:93", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: cyan=9, green=4, orange=9, brown=6, blue=5, white=7, gray=3, pink=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"pink": "The sampled ball is pink", "green": "The sampled ball is green", "gray": "The sampled ball is gray", "orange": "The sampled ball is orange", "brown": "The sampled ball is brown", "blue": "The sampled ball is blue", "white": "The sampled ball is white", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: cyan, brown, gray", "The drawn color is one of: green, blue, pink", "The drawn color is one of: orange, white"]}}, "gold": {}, "gold_probs": {"outcome": {"pink": 0.1568627450980392, "green": 0.0784313725490196, "gray": 0.058823529411764705, "orange": 0.17647058823529413, "brown": 0.11764705882352941, "blue": 0.09803921568627451, "white": 0.13725490196078433, "cyan": 0.17647058823529413}, "focus": {"false": 0.9019607843137255, "true": 0.09803921568627451}, "group": {"0": 0.35294117647058826, "1": 0.3333333333333333, "2": 0.3137254901960784}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"5/51\"],[\"brown\",\"2/17\"],[\"cyan\",\"3/17\"],[\"gray\",\"1/17\"],[\"green\",\"4/51\"],[\"orange\",\"3/17\"],[\"pink\",\"8/51\"],[\"white\",\"7/51\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"brown": 0.01, "gray": 0.03, "cyan": 0.71, "pink": 0.02, "white": 0.01, "blue": 0.01, "green": 0.01, "orange": 0.2}, "focus": {"false": 0.87, "true": 0.13}, "group": {"0": 0.33, "1": 0.05, "2": 0.62}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:96", "state_id": "known_chance_v2:train:96", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: brown=8, orange=6, gray=9, purple=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"brown": "The sampled ball is brown", "gray": "The sampled ball is gray", "purple": "The sampled ball is purple", "orange": "The sampled ball is orange"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a orange ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: brown, purple", "The drawn color is one of: orange", "The drawn color is one of: gray"]}}, "gold": {}, "gold_probs": {"outcome": {"brown": 0.27586206896551724, "gray": 0.3103448275862069, "purple": 0.20689655172413793, "orange": 0.20689655172413793}, "focus": {"false": 0.7931034482758621, "true": 0.20689655172413793}, "group": {"0": 0.48275862068965514, "1": 0.20689655172413793, "2": 0.3103448275862069}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"8/29\"],[\"gray\",\"9/29\"],[\"orange\",\"6/29\"],[\"purple\",\"6/29\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"gray": 0.84, "brown": 0.15, "orange": 0.01, "purple": 0}, "focus": {"false": 0.81, "true": 0.19}, "group": {"0": 0.1, "1": 0.01, "2": 0.89}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:97", "state_id": "known_chance_v2:train:97", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: yellow=1, pink=9, purple=3, blue=4, red=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"purple": "The sampled ball is purple", "yellow": "The sampled ball is yellow", "blue": "The sampled ball is blue", "red": "The sampled ball is red", "pink": "The sampled ball is pink"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: yellow, blue", "The drawn color is one of: pink, red", "The drawn color is one of: purple"]}}, "gold": {}, "gold_probs": {"outcome": {"purple": 0.15, "yellow": 0.05, "blue": 0.2, "red": 0.15, "pink": 0.45}, "focus": {"false": 0.8, "true": 0.2}, "group": {"0": 0.25, "1": 0.6, "2": 0.15}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"1/5\"],[\"pink\",\"9/20\"],[\"purple\",\"3/20\"],[\"red\",\"3/20\"],[\"yellow\",\"1/20\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"yellow": 0.03, "blue": 0, "red": 0, "purple": 0, "pink": 0.97}, "focus": {"false": 0.8, "true": 0.2}, "group": {"0": 0.01, "1": 0.98, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:95", "state_id": "known_chance_v2:train:95", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: blue=6, yellow=7, gray=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"yellow": "The sampled ball is yellow", "gray": "The sampled ball is gray", "blue": "The sampled ball is blue"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a gray ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue", "The drawn color is one of: yellow", "The drawn color is one of: gray"]}}, "gold": {}, "gold_probs": {"outcome": {"yellow": 0.5, "gray": 0.07142857142857142, "blue": 0.42857142857142855}, "focus": {"false": 0.9285714285714286, "true": 0.07142857142857142}, "group": {"0": 0.42857142857142855, "1": 0.5, "2": 0.07142857142857142}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"3/7\"],[\"gray\",\"1/14\"],[\"yellow\",\"1/2\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"gray": 0.02, "yellow": 0.72, "blue": 0.26}, "focus": {"false": 0.94, "true": 0.06}, "group": {"0": 0.04, "1": 0.95, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:94", "state_id": "known_chance_v2:train:94", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: brown=7, white=4, orange=3, yellow=1, green=7, red=4, purple=8, pink=8, black=8, blue=4, cyan=4, gray=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "red": "The sampled ball is red", "black": "The sampled ball is black", "yellow": "The sampled ball is yellow", "orange": "The sampled ball is orange", "white": "The sampled ball is white", "gray": "The sampled ball is gray", "blue": "The sampled ball is blue", "purple": "The sampled ball is purple", "pink": "The sampled ball is pink", "green": "The sampled ball is green", "brown": "The sampled ball is brown"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a orange ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: brown, yellow, purple, blue", "The drawn color is one of: white, green, pink, cyan", "The drawn color is one of: orange, red, black, gray"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.06451612903225806, "red": 0.06451612903225806, "black": 0.12903225806451613, "yellow": 0.016129032258064516, "orange": 0.04838709677419355, "white": 0.06451612903225806, "gray": 0.06451612903225806, "blue": 0.06451612903225806, "purple": 0.12903225806451613, "pink": 0.12903225806451613, "green": 0.11290322580645161, "brown": 0.11290322580645161}, "focus": {"false": 0.9516129032258065, "true": 0.04838709677419355}, "group": {"0": 0.3225806451612903, "1": 0.3709677419354839, "2": 0.3064516129032258}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"4/31\"],[\"blue\",\"2/31\"],[\"brown\",\"7/62\"],[\"cyan\",\"2/31\"],[\"gray\",\"2/31\"],[\"green\",\"7/62\"],[\"orange\",\"3/62\"],[\"pink\",\"4/31\"],[\"purple\",\"4/31\"],[\"red\",\"2/31\"],[\"white\",\"2/31\"],[\"yellow\",\"1/62\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0.01, "green": 0.03, "blue": 0, "purple": 0.27, "pink": 0.07, "white": 0.01, "red": 0.01, "brown": 0.33999999999999997, "black": 0.14, "orange": 0.01, "gray": 0.01, "yellow": 0.1}, "focus": {"false": 0.91, "true": 0.09}, "group": {"0": 0.44, "1": 0.36, "2": 0.2}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:98", "state_id": "known_chance_v2:train:98", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: purple=6, red=4, green=3, blue=1, pink=8, yellow=9, cyan=7, orange=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"yellow": "The sampled ball is yellow", "purple": "The sampled ball is purple", "blue": "The sampled ball is blue", "orange": "The sampled ball is orange", "red": "The sampled ball is red", "cyan": "The sampled ball is cyan", "pink": "The sampled ball is pink", "green": "The sampled ball is green"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: purple, blue, cyan", "The drawn color is one of: red, pink, orange", "The drawn color is one of: green, yellow"]}}, "gold": {}, "gold_probs": {"outcome": {"yellow": 0.20930232558139536, "purple": 0.13953488372093023, "blue": 0.023255813953488372, "orange": 0.11627906976744186, "red": 0.09302325581395349, "cyan": 0.16279069767441862, "pink": 0.18604651162790697, "green": 0.06976744186046512}, "focus": {"false": 0.8372093023255813, "true": 0.16279069767441862}, "group": {"0": 0.32558139534883723, "1": 0.3953488372093023, "2": 0.27906976744186046}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"1/43\"],[\"cyan\",\"7/43\"],[\"green\",\"3/43\"],[\"orange\",\"5/43\"],[\"pink\",\"8/43\"],[\"purple\",\"6/43\"],[\"red\",\"4/43\"],[\"yellow\",\"9/43\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0, "blue": 0.03, "yellow": 0.88, "purple": 0.04, "red": 0, "orange": 0, "pink": 0.04, "cyan": 0.01}, "focus": {"false": 0.86, "true": 0.14}, "group": {"0": 0.07, "1": 0.49, "2": 0.44}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:100", "state_id": "known_chance_v2:train:100", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: blue=6, purple=8, cyan=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"purple": "The sampled ball is purple", "blue": "The sampled ball is blue", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue", "The drawn color is one of: purple", "The drawn color is one of: cyan"]}}, "gold": {}, "gold_probs": {"outcome": {"purple": 0.42105263157894735, "blue": 0.3157894736842105, "cyan": 0.2631578947368421}, "focus": {"false": 0.736842105263158, "true": 0.2631578947368421}, "group": {"0": 0.3157894736842105, "1": 0.42105263157894735, "2": 0.2631578947368421}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"6/19\"],[\"cyan\",\"5/19\"],[\"purple\",\"8/19\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.86, "blue": 0.12, "cyan": 0.02}, "focus": {"false": 0.81, "true": 0.19}, "group": {"0": 0.02, "1": 0.97, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:101", "state_id": "known_chance_v2:train:101", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: gray=8, green=4, purple=8, orange=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "gray": "The sampled ball is gray", "orange": "The sampled ball is orange", "purple": "The sampled ball is purple"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a purple ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: gray, orange", "The drawn color is one of: green", "The drawn color is one of: purple"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.15384615384615385, "gray": 0.3076923076923077, "orange": 0.23076923076923078, "purple": 0.3076923076923077}, "focus": {"false": 0.6923076923076923, "true": 0.3076923076923077}, "group": {"0": 0.5384615384615385, "1": 0.15384615384615385, "2": 0.3076923076923077}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"gray\",\"4/13\"],[\"green\",\"2/13\"],[\"orange\",\"3/13\"],[\"purple\",\"4/13\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0.04, "purple": 0.09, "orange": 0.03, "gray": 0.84}, "focus": {"false": 0.73, "true": 0.27}, "group": {"0": 0.74, "1": 0.01, "2": 0.25}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:102", "state_id": "known_chance_v2:train:102", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: pink=3, blue=5, gray=8, black=5, yellow=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "pink": "The sampled ball is pink", "blue": "The sampled ball is blue", "yellow": "The sampled ball is yellow", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a pink ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: pink, black", "The drawn color is one of: blue, yellow", "The drawn color is one of: gray"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.17857142857142858, "pink": 0.10714285714285714, "blue": 0.17857142857142858, "yellow": 0.25, "gray": 0.2857142857142857}, "focus": {"false": 0.8928571428571429, "true": 0.10714285714285714}, "group": {"0": 0.2857142857142857, "1": 0.4285714285714286, "2": 0.2857142857142857}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"5/28\"],[\"blue\",\"5/28\"],[\"gray\",\"2/7\"],[\"pink\",\"3/28\"],[\"yellow\",\"1/4\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"blue": 0.01, "black": 0, "pink": 0.06, "gray": 0.92, "yellow": 0.01}, "focus": {"false": 0.87, "true": 0.13}, "group": {"0": 0.01, "1": 0.04, "2": 0.95}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:99", "state_id": "known_chance_v2:train:99", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: pink=2, red=5, brown=2, black=3, white=4, gray=3, cyan=4, green=9, orange=2, purple=8, yellow=9, blue=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"orange": "The sampled ball is orange", "pink": "The sampled ball is pink", "white": "The sampled ball is white", "brown": "The sampled ball is brown", "green": "The sampled ball is green", "black": "The sampled ball is black", "blue": "The sampled ball is blue", "yellow": "The sampled ball is yellow", "gray": "The sampled ball is gray", "cyan": "The sampled ball is cyan", "red": "The sampled ball is red", "purple": "The sampled ball is purple"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: pink, black, cyan, purple", "The drawn color is one of: red, white, green, yellow", "The drawn color is one of: brown, gray, orange, blue"]}}, "gold": {}, "gold_probs": {"outcome": {"orange": 0.034482758620689655, "pink": 0.034482758620689655, "white": 0.06896551724137931, "brown": 0.034482758620689655, "green": 0.15517241379310345, "black": 0.05172413793103448, "blue": 0.1206896551724138, "yellow": 0.15517241379310345, "gray": 0.05172413793103448, "cyan": 0.06896551724137931, "red": 0.08620689655172414, "purple": 0.13793103448275862}, "focus": {"false": 0.9310344827586207, "true": 0.06896551724137931}, "group": {"0": 0.29310344827586204, "1": 0.4655172413793104, "2": 0.2413793103448276}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"3/58\"],[\"blue\",\"7/58\"],[\"brown\",\"1/29\"],[\"cyan\",\"2/29\"],[\"gray\",\"3/58\"],[\"green\",\"9/58\"],[\"orange\",\"1/29\"],[\"pink\",\"1/29\"],[\"purple\",\"4/29\"],[\"red\",\"5/58\"],[\"white\",\"2/29\"],[\"yellow\",\"9/58\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.01, "cyan": 0.01, "white": 0, "pink": 0.02, "yellow": 0.25, "gray": 0, "blue": 0.01, "brown": 0, "red": 0.04, "orange": 0, "green": 0.65, "black": 0.01}, "focus": {"false": 0.87, "true": 0.13}, "group": {"0": 0.01, "1": 0.98, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:103", "state_id": "known_chance_v2:train:103", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: orange=8, blue=5, red=3, yellow=8, cyan=1, black=9, white=8, purple=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"orange": "The sampled ball is orange", "yellow": "The sampled ball is yellow", "red": "The sampled ball is red", "cyan": "The sampled ball is cyan", "purple": "The sampled ball is purple", "white": "The sampled ball is white", "blue": "The sampled ball is blue", "black": "The sampled ball is black"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: orange, yellow, white", "The drawn color is one of: blue, cyan, purple", "The drawn color is one of: red, black"]}}, "gold": {}, "gold_probs": {"outcome": {"orange": 0.18181818181818182, "yellow": 0.18181818181818182, "red": 0.06818181818181818, "cyan": 0.022727272727272728, "purple": 0.045454545454545456, "white": 0.18181818181818182, "blue": 0.11363636363636363, "black": 0.20454545454545456}, "focus": {"false": 0.8181818181818181, "true": 0.18181818181818182}, "group": {"0": 0.5454545454545454, "1": 0.18181818181818182, "2": 0.2727272727272727}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"9/44\"],[\"blue\",\"5/44\"],[\"cyan\",\"1/44\"],[\"orange\",\"2/11\"],[\"purple\",\"1/22\"],[\"red\",\"3/44\"],[\"white\",\"2/11\"],[\"yellow\",\"2/11\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0, "cyan": 0.02, "black": 0.9400000000000001, "yellow": 0, "orange": 0.03, "blue": 0, "purple": 0, "red": 0}, "focus": {"false": 0.83, "true": 0.17}, "group": {"0": 0.85, "1": 0.01, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:104", "state_id": "known_chance_v2:train:104", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: gray=8, cyan=8, red=3, yellow=6, black=6, pink=5, white=4, brown=3, green=2, purple=4, orange=2, blue=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "green": "The sampled ball is green", "black": "The sampled ball is black", "cyan": "The sampled ball is cyan", "white": "The sampled ball is white", "pink": "The sampled ball is pink", "gray": "The sampled ball is gray", "yellow": "The sampled ball is yellow", "brown": "The sampled ball is brown", "orange": "The sampled ball is orange", "purple": "The sampled ball is purple", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a red ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: gray, yellow, white, purple", "The drawn color is one of: cyan, black, brown, orange", "The drawn color is one of: red, pink, green, blue"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.03773584905660377, "green": 0.03773584905660377, "black": 0.11320754716981132, "cyan": 0.1509433962264151, "white": 0.07547169811320754, "pink": 0.09433962264150944, "gray": 0.1509433962264151, "yellow": 0.11320754716981132, "brown": 0.05660377358490566, "orange": 0.03773584905660377, "purple": 0.07547169811320754, "red": 0.05660377358490566}, "focus": {"false": 0.9433962264150944, "true": 0.05660377358490566}, "group": {"0": 0.41509433962264153, "1": 0.3584905660377358, "2": 0.22641509433962265}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"6/53\"],[\"blue\",\"2/53\"],[\"brown\",\"3/53\"],[\"cyan\",\"8/53\"],[\"gray\",\"8/53\"],[\"green\",\"2/53\"],[\"orange\",\"2/53\"],[\"pink\",\"5/53\"],[\"purple\",\"4/53\"],[\"red\",\"3/53\"],[\"white\",\"4/53\"],[\"yellow\",\"6/53\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.01, "pink": 0, "yellow": 0.02, "white": 0, "gray": 0.84, "blue": 0.01, "red": 0.01, "orange": 0, "cyan": 0.1, "black": 0.01, "green": 0, "brown": 0}, "focus": {"false": 0.88, "true": 0.12}, "group": {"0": 0.8099999999999999, "1": 0.16, "2": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:105", "state_id": "known_chance_v2:train:105", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: black=4, green=4, cyan=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "cyan": "The sampled ball is cyan", "black": "The sampled ball is black"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a green ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: black", "The drawn color is one of: green", "The drawn color is one of: cyan"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.2857142857142857, "cyan": 0.42857142857142855, "black": 0.2857142857142857}, "focus": {"false": 0.7142857142857143, "true": 0.2857142857142857}, "group": {"0": 0.2857142857142857, "1": 0.2857142857142857, "2": 0.42857142857142855}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"2/7\"],[\"cyan\",\"3/7\"],[\"green\",\"2/7\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0.02, "black": 0.14, "cyan": 0.84}, "focus": {"false": 0.71, "true": 0.29}, "group": {"0": 0.03, "1": 0.01, "2": 0.96}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:106", "state_id": "known_chance_v2:train:106", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: green=2, white=2, gray=9, yellow=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"yellow": "The sampled ball is yellow", "white": "The sampled ball is white", "gray": "The sampled ball is gray", "green": "The sampled ball is green"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: green, yellow", "The drawn color is one of: white", "The drawn color is one of: gray"]}}, "gold": {}, "gold_probs": {"outcome": {"yellow": 0.1875, "white": 0.125, "gray": 0.5625, "green": 0.125}, "focus": {"false": 0.8125, "true": 0.1875}, "group": {"0": 0.3125, "1": 0.125, "2": 0.5625}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"gray\",\"9/16\"],[\"green\",\"1/8\"],[\"white\",\"1/8\"],[\"yellow\",\"3/16\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"yellow": 0, "green": 0.01, "white": 0, "gray": 0.99}, "focus": {"false": 0.86, "true": 0.14}, "group": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:107", "state_id": "known_chance_v2:train:107", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: blue=9, white=3, cyan=5, yellow=6, green=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "white": "The sampled ball is white", "blue": "The sampled ball is blue", "cyan": "The sampled ball is cyan", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue, yellow", "The drawn color is one of: white, green", "The drawn color is one of: cyan"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.041666666666666664, "white": 0.125, "blue": 0.375, "cyan": 0.20833333333333334, "yellow": 0.25}, "focus": {"false": 0.7916666666666666, "true": 0.20833333333333334}, "group": {"0": 0.625, "1": 0.16666666666666666, "2": 0.20833333333333334}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"3/8\"],[\"cyan\",\"5/24\"],[\"green\",\"1/24\"],[\"white\",\"1/8\"],[\"yellow\",\"1/4\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0, "yellow": 0.01, "cyan": 0.01, "blue": 0.97, "green": 0.01}, "focus": {"false": 0.8200000000000001, "true": 0.18}, "group": {"0": 0.99, "1": 0, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:109", "state_id": "known_chance_v2:train:109", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: green=4, pink=4, white=3, orange=1, black=2, gray=9, blue=4, purple=3, red=4, cyan=5, brown=5, yellow=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"purple": "The sampled ball is purple", "yellow": "The sampled ball is yellow", "red": "The sampled ball is red", "white": "The sampled ball is white", "blue": "The sampled ball is blue", "black": "The sampled ball is black", "orange": "The sampled ball is orange", "pink": "The sampled ball is pink", "brown": "The sampled ball is brown", "green": "The sampled ball is green", "cyan": "The sampled ball is cyan", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: green, orange, blue, cyan", "The drawn color is one of: pink, black, purple, brown", "The drawn color is one of: white, gray, red, yellow"]}}, "gold": {}, "gold_probs": {"outcome": {"purple": 0.06382978723404255, "yellow": 0.06382978723404255, "red": 0.0851063829787234, "white": 0.06382978723404255, "blue": 0.0851063829787234, "black": 0.0425531914893617, "orange": 0.02127659574468085, "pink": 0.0851063829787234, "brown": 0.10638297872340426, "green": 0.0851063829787234, "cyan": 0.10638297872340426, "gray": 0.19148936170212766}, "focus": {"false": 0.9574468085106383, "true": 0.0425531914893617}, "group": {"0": 0.2978723404255319, "1": 0.2978723404255319, "2": 0.40425531914893614}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"2/47\"],[\"blue\",\"4/47\"],[\"brown\",\"5/47\"],[\"cyan\",\"5/47\"],[\"gray\",\"9/47\"],[\"green\",\"4/47\"],[\"orange\",\"1/47\"],[\"pink\",\"4/47\"],[\"purple\",\"3/47\"],[\"red\",\"4/47\"],[\"white\",\"3/47\"],[\"yellow\",\"3/47\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0.02, "blue": 0, "white": 0, "gray": 0.97, "yellow": 0, "purple": 0, "red": 0, "orange": 0.01, "pink": 0, "cyan": 0, "brown": 0, "black": 0}, "focus": {"false": 0.89, "true": 0.11}, "group": {"0": 0.06, "1": 0.03, "2": 0.91}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:108", "state_id": "known_chance_v2:train:108", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: red=9, cyan=6, gray=6, orange=9, pink=8, yellow=9, purple=7, white=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "yellow": "The sampled ball is yellow", "red": "The sampled ball is red", "gray": "The sampled ball is gray", "pink": "The sampled ball is pink", "purple": "The sampled ball is purple", "white": "The sampled ball is white", "orange": "The sampled ball is orange"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a orange ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: red, orange, purple", "The drawn color is one of: cyan, pink, white", "The drawn color is one of: gray, yellow"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.10526315789473684, "yellow": 0.15789473684210525, "red": 0.15789473684210525, "gray": 0.10526315789473684, "pink": 0.14035087719298245, "purple": 0.12280701754385964, "white": 0.05263157894736842, "orange": 0.15789473684210525}, "focus": {"false": 0.8421052631578947, "true": 0.15789473684210525}, "group": {"0": 0.43859649122807015, "1": 0.2982456140350877, "2": 0.2631578947368421}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"2/19\"],[\"gray\",\"2/19\"],[\"orange\",\"3/19\"],[\"pink\",\"8/57\"],[\"purple\",\"7/57\"],[\"red\",\"3/19\"],[\"white\",\"1/19\"],[\"yellow\",\"3/19\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0.01, "purple": 0.01, "pink": 0.03, "white": 0.11, "red": 0.76, "orange": 0.03, "gray": 0.01, "yellow": 0.04}, "focus": {"false": 0.81, "true": 0.19}, "group": {"0": 0.93, "1": 0.03, "2": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:110", "state_id": "known_chance_v2:train:110", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: brown=1, yellow=1, cyan=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "brown": "The sampled ball is brown", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a brown ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: brown", "The drawn color is one of: yellow", "The drawn color is one of: cyan"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.8, "brown": 0.1, "yellow": 0.1}, "focus": {"false": 0.9, "true": 0.1}, "group": {"0": 0.1, "1": 0.1, "2": 0.8}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"1/10\"],[\"cyan\",\"4/5\"],[\"yellow\",\"1/10\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 1, "brown": 0, "yellow": 0}, "focus": {"false": 0.94, "true": 0.06}, "group": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:113", "state_id": "known_chance_v2:train:113", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: cyan=1, white=2, brown=3, yellow=5, gray=6, red=6, blue=1, green=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"white": "The sampled ball is white", "cyan": "The sampled ball is cyan", "blue": "The sampled ball is blue", "green": "The sampled ball is green", "gray": "The sampled ball is gray", "yellow": "The sampled ball is yellow", "brown": "The sampled ball is brown", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: cyan, yellow, blue", "The drawn color is one of: white, gray, green", "The drawn color is one of: brown, red"]}}, "gold": {}, "gold_probs": {"outcome": {"white": 0.08, "cyan": 0.04, "blue": 0.04, "green": 0.04, "gray": 0.24, "yellow": 0.2, "brown": 0.12, "red": 0.24}, "focus": {"false": 0.96, "true": 0.04}, "group": {"0": 0.28, "1": 0.36, "2": 0.36}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"1/25\"],[\"brown\",\"3/25\"],[\"cyan\",\"1/25\"],[\"gray\",\"6/25\"],[\"green\",\"1/25\"],[\"red\",\"6/25\"],[\"white\",\"2/25\"],[\"yellow\",\"1/5\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0.01, "red": 0.12, "cyan": 0.02, "gray": 0.52, "brown": 0, "yellow": 0.32, "blue": 0, "green": 0.01}, "focus": {"false": 0.92, "true": 0.08}, "group": {"0": 0.24, "1": 0.32, "2": 0.44}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:114", "state_id": "known_chance_v2:train:114", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: brown=5, cyan=2, black=7, red=2, blue=8, pink=8, green=6, yellow=7, gray=9, white=4, purple=8, orange=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "cyan": "The sampled ball is cyan", "purple": "The sampled ball is purple", "pink": "The sampled ball is pink", "gray": "The sampled ball is gray", "white": "The sampled ball is white", "red": "The sampled ball is red", "black": "The sampled ball is black", "yellow": "The sampled ball is yellow", "green": "The sampled ball is green", "orange": "The sampled ball is orange", "brown": "The sampled ball is brown"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a white ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: brown, red, green, white", "The drawn color is one of: cyan, blue, yellow, purple", "The drawn color is one of: black, pink, gray, orange"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.1111111111111111, "cyan": 0.027777777777777776, "purple": 0.1111111111111111, "pink": 0.1111111111111111, "gray": 0.125, "white": 0.05555555555555555, "red": 0.027777777777777776, "black": 0.09722222222222222, "yellow": 0.09722222222222222, "green": 0.08333333333333333, "orange": 0.08333333333333333, "brown": 0.06944444444444445}, "focus": {"false": 0.9444444444444444, "true": 0.05555555555555555}, "group": {"0": 0.2361111111111111, "1": 0.3472222222222222, "2": 0.41666666666666663}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"7/72\"],[\"blue\",\"1/9\"],[\"brown\",\"5/72\"],[\"cyan\",\"1/36\"],[\"gray\",\"1/8\"],[\"green\",\"1/12\"],[\"orange\",\"1/12\"],[\"pink\",\"1/9\"],[\"purple\",\"1/9\"],[\"red\",\"1/36\"],[\"white\",\"1/18\"],[\"yellow\",\"7/72\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.01, "pink": 0, "green": 0, "orange": 0, "blue": 0.01, "cyan": 0.01, "red": 0, "white": 0.01, "yellow": 0, "brown": 0.02, "black": 0.01, "gray": 0.93}, "focus": {"false": 0.89, "true": 0.11}, "group": {"0": 0.02, "1": 0.09, "2": 0.89}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:111", "state_id": "known_chance_v2:train:111", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: green=7, brown=8, blue=8, black=9. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"brown": "The sampled ball is brown", "green": "The sampled ball is green", "black": "The sampled ball is black", "blue": "The sampled ball is blue"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: green, black", "The drawn color is one of: brown", "The drawn color is one of: blue"]}}, "gold": {}, "gold_probs": {"outcome": {"brown": 0.25, "green": 0.21875, "black": 0.28125, "blue": 0.25}, "focus": {"false": 0.75, "true": 0.25}, "group": {"0": 0.5, "1": 0.25, "2": 0.25}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"9/32\"],[\"blue\",\"1/4\"],[\"brown\",\"1/4\"],[\"green\",\"7/32\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"brown": 0.03, "blue": 0.02, "green": 0.23, "black": 0.72}, "focus": {"false": 0.77, "true": 0.23}, "group": {"0": 0.83, "1": 0.1, "2": 0.07}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:115", "state_id": "known_chance_v2:train:115", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: orange=6, yellow=1, green=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"orange": "The sampled ball is orange", "yellow": "The sampled ball is yellow", "green": "The sampled ball is green"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a green ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: orange", "The drawn color is one of: yellow", "The drawn color is one of: green"]}}, "gold": {}, "gold_probs": {"outcome": {"orange": 0.5, "yellow": 0.08333333333333333, "green": 0.4166666666666667}, "focus": {"false": 0.5833333333333333, "true": 0.4166666666666667}, "group": {"0": 0.5, "1": 0.08333333333333333, "2": 0.4166666666666667}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"green\",\"5/12\"],[\"orange\",\"1/2\"],[\"yellow\",\"1/12\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"yellow": 0.02, "orange": 0.95, "green": 0.03}, "focus": {"false": 0.71, "true": 0.29}, "group": {"0": 0.95, "1": 0.01, "2": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:112", "state_id": "known_chance_v2:train:112", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: red=4, white=8, black=8, pink=7, blue=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "pink": "The sampled ball is pink", "white": "The sampled ball is white", "black": "The sampled ball is black", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: red, pink", "The drawn color is one of: white, blue", "The drawn color is one of: black"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.03571428571428571, "pink": 0.25, "white": 0.2857142857142857, "black": 0.2857142857142857, "red": 0.14285714285714285}, "focus": {"false": 0.9642857142857143, "true": 0.03571428571428571}, "group": {"0": 0.39285714285714285, "1": 0.3214285714285714, "2": 0.2857142857142857}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"2/7\"],[\"blue\",\"1/28\"],[\"pink\",\"1/4\"],[\"red\",\"1/7\"],[\"white\",\"2/7\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.04, "black": 0.19, "red": 0.03, "blue": 0.04, "white": 0.7}, "focus": {"false": 0.95, "true": 0.05}, "group": {"0": 0.15, "1": 0.57, "2": 0.28}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:116", "state_id": "known_chance_v2:train:116", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: blue=4, purple=2, red=1, gray=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"purple": "The sampled ball is purple", "blue": "The sampled ball is blue", "gray": "The sampled ball is gray", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a gray ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue, gray", "The drawn color is one of: purple", "The drawn color is one of: red"]}}, "gold": {}, "gold_probs": {"outcome": {"purple": 0.14285714285714285, "blue": 0.2857142857142857, "gray": 0.5, "red": 0.07142857142857142}, "focus": {"false": 0.5, "true": 0.5}, "group": {"0": 0.7857142857142857, "1": 0.14285714285714285, "2": 0.07142857142857142}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"2/7\"],[\"gray\",\"1/2\"],[\"purple\",\"1/7\"],[\"red\",\"1/14\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"blue": 0.07, "purple": 0, "gray": 0.92, "red": 0.01}, "focus": {"false": 0.6599999999999999, "true": 0.34}, "group": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:117", "state_id": "known_chance_v2:train:117", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: pink=2, cyan=7, yellow=4, green=2, white=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "cyan": "The sampled ball is cyan", "white": "The sampled ball is white", "yellow": "The sampled ball is yellow", "pink": "The sampled ball is pink"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a green ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: pink, green", "The drawn color is one of: cyan, white", "The drawn color is one of: yellow"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.1111111111111111, "cyan": 0.3888888888888889, "white": 0.16666666666666666, "yellow": 0.2222222222222222, "pink": 0.1111111111111111}, "focus": {"false": 0.8888888888888888, "true": 0.1111111111111111}, "group": {"0": 0.2222222222222222, "1": 0.5555555555555556, "2": 0.2222222222222222}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"7/18\"],[\"green\",\"1/9\"],[\"pink\",\"1/9\"],[\"white\",\"1/6\"],[\"yellow\",\"2/9\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0.99, "green": 0, "yellow": 0, "white": 0, "pink": 0.01}, "focus": {"false": 0.85, "true": 0.15}, "group": {"0": 0.02, "1": 0.96, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:118", "state_id": "known_chance_v2:train:118", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: gray=2, pink=5, purple=6, white=9, blue=9, cyan=8, green=9, brown=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"brown": "The sampled ball is brown", "blue": "The sampled ball is blue", "white": "The sampled ball is white", "gray": "The sampled ball is gray", "purple": "The sampled ball is purple", "pink": "The sampled ball is pink", "cyan": "The sampled ball is cyan", "green": "The sampled ball is green"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a gray ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: gray, white, green", "The drawn color is one of: pink, blue, brown", "The drawn color is one of: purple, cyan"]}}, "gold": {}, "gold_probs": {"outcome": {"brown": 0.04, "blue": 0.18, "white": 0.18, "gray": 0.04, "purple": 0.12, "pink": 0.1, "cyan": 0.16, "green": 0.18}, "focus": {"false": 0.96, "true": 0.04}, "group": {"0": 0.39999999999999997, "1": 0.32, "2": 0.28}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"9/50\"],[\"brown\",\"1/25\"],[\"cyan\",\"4/25\"],[\"gray\",\"1/25\"],[\"green\",\"9/50\"],[\"pink\",\"1/10\"],[\"purple\",\"3/25\"],[\"white\",\"9/50\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.02, "pink": 0.01, "cyan": 0.04, "brown": 0.01, "gray": 0.09, "white": 0.68, "blue": 0.08, "green": 0.07}, "focus": {"false": 0.91, "true": 0.09}, "group": {"0": 0.8, "1": 0.09, "2": 0.11}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:120", "state_id": "known_chance_v2:train:120", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: orange=2, yellow=2, red=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"orange": "The sampled ball is orange", "yellow": "The sampled ball is yellow", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a orange ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: orange", "The drawn color is one of: yellow", "The drawn color is one of: red"]}}, "gold": {}, "gold_probs": {"outcome": {"orange": 0.2222222222222222, "yellow": 0.2222222222222222, "red": 0.5555555555555556}, "focus": {"false": 0.7777777777777778, "true": 0.2222222222222222}, "group": {"0": 0.2222222222222222, "1": 0.2222222222222222, "2": 0.5555555555555556}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"orange\",\"2/9\"],[\"red\",\"5/9\"],[\"yellow\",\"2/9\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"yellow": 0, "orange": 0.03, "red": 0.97}, "focus": {"false": 0.81, "true": 0.19}, "group": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:119", "state_id": "known_chance_v2:train:119", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: cyan=4, black=5, white=6, pink=1, purple=6, orange=8, yellow=2, brown=3, green=6, red=2, blue=7, gray=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "orange": "The sampled ball is orange", "cyan": "The sampled ball is cyan", "white": "The sampled ball is white", "yellow": "The sampled ball is yellow", "green": "The sampled ball is green", "brown": "The sampled ball is brown", "purple": "The sampled ball is purple", "red": "The sampled ball is red", "gray": "The sampled ball is gray", "blue": "The sampled ball is blue", "pink": "The sampled ball is pink"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: cyan, pink, yellow, red", "The drawn color is one of: black, purple, brown, blue", "The drawn color is one of: white, orange, green, gray"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.09615384615384616, "orange": 0.15384615384615385, "cyan": 0.07692307692307693, "white": 0.11538461538461539, "yellow": 0.038461538461538464, "green": 0.11538461538461539, "brown": 0.057692307692307696, "purple": 0.11538461538461539, "red": 0.038461538461538464, "gray": 0.038461538461538464, "blue": 0.1346153846153846, "pink": 0.019230769230769232}, "focus": {"false": 0.9230769230769231, "true": 0.07692307692307693}, "group": {"0": 0.17307692307692307, "1": 0.40384615384615385, "2": 0.42307692307692313}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"5/52\"],[\"blue\",\"7/52\"],[\"brown\",\"3/52\"],[\"cyan\",\"1/13\"],[\"gray\",\"1/26\"],[\"green\",\"3/26\"],[\"orange\",\"2/13\"],[\"pink\",\"1/52\"],[\"purple\",\"3/26\"],[\"red\",\"1/26\"],[\"white\",\"3/26\"],[\"yellow\",\"1/26\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.03, "black": 0.01, "cyan": 0.02, "brown": 0, "purple": 0.01, "green": 0.01, "gray": 0, "blue": 0.02, "white": 0.02, "red": 0, "yellow": 0, "orange": 0.88}, "focus": {"false": 0.89, "true": 0.11}, "group": {"0": 0, "1": 0.11, "2": 0.89}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:121", "state_id": "known_chance_v2:train:121", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: gray=3, blue=1, black=3, white=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "gray": "The sampled ball is gray", "black": "The sampled ball is black", "white": "The sampled ball is white"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: gray, white", "The drawn color is one of: blue", "The drawn color is one of: black"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.08333333333333333, "gray": 0.25, "black": 0.25, "white": 0.4166666666666667}, "focus": {"false": 0.9166666666666666, "true": 0.08333333333333333}, "group": {"0": 0.6666666666666667, "1": 0.08333333333333333, "2": 0.25}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/4\"],[\"blue\",\"1/12\"],[\"gray\",\"1/4\"],[\"white\",\"5/12\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"blue": 0.01, "black": 0.01, "gray": 0.05, "white": 0.93}, "focus": {"false": 0.87, "true": 0.13}, "group": {"0": 0.98, "1": 0.01, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:122", "state_id": "known_chance_v2:train:122", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: purple=9, red=3, orange=1, white=3, cyan=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"purple": "The sampled ball is purple", "white": "The sampled ball is white", "orange": "The sampled ball is orange", "cyan": "The sampled ball is cyan", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: purple, white", "The drawn color is one of: red, cyan", "The drawn color is one of: orange"]}}, "gold": {}, "gold_probs": {"outcome": {"purple": 0.5, "white": 0.16666666666666666, "orange": 0.05555555555555555, "cyan": 0.1111111111111111, "red": 0.16666666666666666}, "focus": {"false": 0.8888888888888888, "true": 0.1111111111111111}, "group": {"0": 0.6666666666666666, "1": 0.2777777777777778, "2": 0.05555555555555555}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"1/9\"],[\"orange\",\"1/18\"],[\"purple\",\"1/2\"],[\"red\",\"1/6\"],[\"white\",\"1/6\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"red": 0, "cyan": 0, "purple": 0.99, "orange": 0.01, "white": 0}, "focus": {"false": 0.88, "true": 0.12}, "group": {"0": 0.98, "1": 0.01, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:123", "state_id": "known_chance_v2:train:123", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: white=4, brown=8, gray=7, purple=9, cyan=3, green=3, black=4, blue=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "black": "The sampled ball is black", "brown": "The sampled ball is brown", "white": "The sampled ball is white", "blue": "The sampled ball is blue", "gray": "The sampled ball is gray", "cyan": "The sampled ball is cyan", "purple": "The sampled ball is purple"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white, purple, black", "The drawn color is one of: brown, cyan, blue", "The drawn color is one of: gray, green"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.07317073170731707, "black": 0.0975609756097561, "brown": 0.1951219512195122, "white": 0.0975609756097561, "blue": 0.07317073170731707, "gray": 0.17073170731707318, "cyan": 0.07317073170731707, "purple": 0.21951219512195122}, "focus": {"false": 0.926829268292683, "true": 0.07317073170731707}, "group": {"0": 0.41463414634146345, "1": 0.3414634146341463, "2": 0.24390243902439024}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"4/41\"],[\"blue\",\"3/41\"],[\"brown\",\"8/41\"],[\"cyan\",\"3/41\"],[\"gray\",\"7/41\"],[\"green\",\"3/41\"],[\"purple\",\"9/41\"],[\"white\",\"4/41\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.83, "green": 0, "white": 0.05, "gray": 0.01, "brown": 0.1, "cyan": 0, "blue": 0.01, "black": 0}, "focus": {"false": 0.88, "true": 0.12}, "group": {"0": 0.47, "1": 0.51, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:124", "state_id": "known_chance_v2:train:124", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: red=3, white=9, brown=7, gray=2, black=8, green=8, purple=4, pink=4, blue=5, yellow=9, orange=1, cyan=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"gray": "The sampled ball is gray", "black": "The sampled ball is black", "white": "The sampled ball is white", "purple": "The sampled ball is purple", "yellow": "The sampled ball is yellow", "brown": "The sampled ball is brown", "orange": "The sampled ball is orange", "cyan": "The sampled ball is cyan", "blue": "The sampled ball is blue", "red": "The sampled ball is red", "green": "The sampled ball is green", "pink": "The sampled ball is pink"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a purple ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: red, gray, purple, yellow", "The drawn color is one of: white, black, pink, orange", "The drawn color is one of: brown, green, blue, cyan"]}}, "gold": {}, "gold_probs": {"outcome": {"gray": 0.03278688524590164, "black": 0.13114754098360656, "white": 0.14754098360655737, "purple": 0.06557377049180328, "yellow": 0.14754098360655737, "brown": 0.11475409836065574, "orange": 0.01639344262295082, "cyan": 0.01639344262295082, "blue": 0.08196721311475409, "red": 0.04918032786885246, "green": 0.13114754098360656, "pink": 0.06557377049180328}, "focus": {"false": 0.9344262295081968, "true": 0.06557377049180328}, "group": {"0": 0.29508196721311475, "1": 0.36065573770491804, "2": 0.3442622950819672}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"8/61\"],[\"blue\",\"5/61\"],[\"brown\",\"7/61\"],[\"cyan\",\"1/61\"],[\"gray\",\"2/61\"],[\"green\",\"8/61\"],[\"orange\",\"1/61\"],[\"pink\",\"4/61\"],[\"purple\",\"4/61\"],[\"red\",\"3/61\"],[\"white\",\"9/61\"],[\"yellow\",\"9/61\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"black": 0.02, "yellow": 0.1, "orange": 0.02, "green": 0.01, "blue": 0.01, "pink": 0, "brown": 0.03, "gray": 0, "white": 0.71, "cyan": 0.02, "red": 0.08, "purple": 0}, "focus": {"false": 0.9, "true": 0.1}, "group": {"0": 0.14, "1": 0.44, "2": 0.42}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:125", "state_id": "known_chance_v2:train:125", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: blue=2, orange=6, cyan=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "orange": "The sampled ball is orange", "blue": "The sampled ball is blue"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue", "The drawn color is one of: orange", "The drawn color is one of: cyan"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.42857142857142855, "orange": 0.42857142857142855, "blue": 0.14285714285714285}, "focus": {"false": 0.5714285714285714, "true": 0.42857142857142855}, "group": {"0": 0.14285714285714285, "1": 0.42857142857142855, "2": 0.42857142857142855}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"1/7\"],[\"cyan\",\"3/7\"],[\"orange\",\"3/7\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"orange": 0.77, "blue": 0.15, "cyan": 0.08}, "focus": {"false": 0.6799999999999999, "true": 0.32}, "group": {"0": 0.03, "1": 0.87, "2": 0.1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:126", "state_id": "known_chance_v2:train:126", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: brown=8, green=7, white=4, orange=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "white": "The sampled ball is white", "brown": "The sampled ball is brown", "orange": "The sampled ball is orange"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a green ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: brown, orange", "The drawn color is one of: green", "The drawn color is one of: white"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.2692307692307692, "white": 0.15384615384615385, "brown": 0.3076923076923077, "orange": 0.2692307692307692}, "focus": {"false": 0.7307692307692308, "true": 0.2692307692307692}, "group": {"0": 0.5769230769230769, "1": 0.2692307692307692, "2": 0.15384615384615385}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"4/13\"],[\"green\",\"7/26\"],[\"orange\",\"7/26\"],[\"white\",\"2/13\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"orange": 0.01, "white": 0.01, "brown": 0.96, "green": 0.02}, "focus": {"false": 0.75, "true": 0.25}, "group": {"0": 0.96, "1": 0.03, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:127", "state_id": "known_chance_v2:train:127", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: cyan=9, green=1, orange=1, red=8, yellow=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "yellow": "The sampled ball is yellow", "orange": "The sampled ball is orange", "green": "The sampled ball is green", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a red ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: cyan, red", "The drawn color is one of: green, yellow", "The drawn color is one of: orange"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.42857142857142855, "yellow": 0.09523809523809523, "orange": 0.047619047619047616, "green": 0.047619047619047616, "red": 0.38095238095238093}, "focus": {"false": 0.6190476190476191, "true": 0.38095238095238093}, "group": {"0": 0.8095238095238095, "1": 0.14285714285714285, "2": 0.047619047619047616}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"3/7\"],[\"green\",\"1/21\"],[\"orange\",\"1/21\"],[\"red\",\"8/21\"],[\"yellow\",\"2/21\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"orange": 0, "green": 0.01, "cyan": 0.8099999999999999, "yellow": 0.01, "red": 0.17}, "focus": {"false": 0.74, "true": 0.26}, "group": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:128", "state_id": "known_chance_v2:train:128", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: red=1, black=1, green=9, purple=1, yellow=5, pink=5, orange=4, white=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"orange": "The sampled ball is orange", "pink": "The sampled ball is pink", "white": "The sampled ball is white", "red": "The sampled ball is red", "green": "The sampled ball is green", "purple": "The sampled ball is purple", "black": "The sampled ball is black", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: red, purple, orange", "The drawn color is one of: black, yellow, white", "The drawn color is one of: green, pink"]}}, "gold": {}, "gold_probs": {"outcome": {"orange": 0.13793103448275862, "pink": 0.1724137931034483, "white": 0.10344827586206896, "red": 0.034482758620689655, "green": 0.3103448275862069, "purple": 0.034482758620689655, "black": 0.034482758620689655, "yellow": 0.1724137931034483}, "focus": {"false": 0.9655172413793104, "true": 0.034482758620689655}, "group": {"0": 0.20689655172413793, "1": 0.3103448275862069, "2": 0.4827586206896552}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/29\"],[\"green\",\"9/29\"],[\"orange\",\"4/29\"],[\"pink\",\"5/29\"],[\"purple\",\"1/29\"],[\"red\",\"1/29\"],[\"white\",\"3/29\"],[\"yellow\",\"5/29\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0.99, "orange": 0, "white": 0, "purple": 0, "pink": 0, "red": 0, "yellow": 0.01, "black": 0}, "focus": {"false": 0.9299999999999999, "true": 0.07}, "group": {"0": 0, "1": 0.02, "2": 0.98}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:129", "state_id": "known_chance_v2:train:129", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: red=4, yellow=8, brown=5, cyan=8, blue=4, green=5, pink=2, gray=4, orange=6, black=6, purple=4, white=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "yellow": "The sampled ball is yellow", "pink": "The sampled ball is pink", "cyan": "The sampled ball is cyan", "purple": "The sampled ball is purple", "black": "The sampled ball is black", "red": "The sampled ball is red", "brown": "The sampled ball is brown", "white": "The sampled ball is white", "orange": "The sampled ball is orange", "gray": "The sampled ball is gray", "blue": "The sampled ball is blue"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a brown ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: red, cyan, pink, black", "The drawn color is one of: yellow, blue, gray, purple", "The drawn color is one of: brown, green, orange, white"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.0847457627118644, "yellow": 0.13559322033898305, "pink": 0.03389830508474576, "cyan": 0.13559322033898305, "purple": 0.06779661016949153, "black": 0.1016949152542373, "red": 0.06779661016949153, "brown": 0.0847457627118644, "white": 0.05084745762711865, "orange": 0.1016949152542373, "gray": 0.06779661016949153, "blue": 0.06779661016949153}, "focus": {"false": 0.9152542372881356, "true": 0.0847457627118644}, "group": {"0": 0.3389830508474576, "1": 0.3389830508474576, "2": 0.3220338983050848}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"6/59\"],[\"blue\",\"4/59\"],[\"brown\",\"5/59\"],[\"cyan\",\"8/59\"],[\"gray\",\"4/59\"],[\"green\",\"5/59\"],[\"orange\",\"6/59\"],[\"pink\",\"2/59\"],[\"purple\",\"4/59\"],[\"red\",\"4/59\"],[\"white\",\"3/59\"],[\"yellow\",\"8/59\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0.01, "orange": 0.02, "blue": 0.01, "brown": 0.01, "yellow": 0.55, "gray": 0.01, "black": 0.01, "pink": 0.04, "cyan": 0.12, "purple": 0.01, "green": 0.01, "red": 0.2}, "focus": {"false": 0.84, "true": 0.16}, "group": {"0": 0.14, "1": 0.53, "2": 0.33}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:130", "state_id": "known_chance_v2:train:130", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: gray=5, pink=3, red=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"gray": "The sampled ball is gray", "red": "The sampled ball is red", "pink": "The sampled ball is pink"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a red ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: gray", "The drawn color is one of: pink", "The drawn color is one of: red"]}}, "gold": {}, "gold_probs": {"outcome": {"gray": 0.38461538461538464, "red": 0.38461538461538464, "pink": 0.23076923076923078}, "focus": {"false": 0.6153846153846154, "true": 0.38461538461538464}, "group": {"0": 0.38461538461538464, "1": 0.23076923076923078, "2": 0.38461538461538464}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"gray\",\"5/13\"],[\"pink\",\"3/13\"],[\"red\",\"5/13\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.04, "gray": 0.89, "red": 0.07}, "focus": {"false": 0.65, "true": 0.35}, "group": {"0": 0.81, "1": 0.06, "2": 0.13}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:131", "state_id": "known_chance_v2:train:131", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: white=3, orange=9, brown=5, black=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"orange": "The sampled ball is orange", "brown": "The sampled ball is brown", "black": "The sampled ball is black", "white": "The sampled ball is white"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a orange ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white, black", "The drawn color is one of: orange", "The drawn color is one of: brown"]}}, "gold": {}, "gold_probs": {"outcome": {"orange": 0.42857142857142855, "brown": 0.23809523809523808, "black": 0.19047619047619047, "white": 0.14285714285714285}, "focus": {"false": 0.5714285714285714, "true": 0.42857142857142855}, "group": {"0": 0.3333333333333333, "1": 0.42857142857142855, "2": 0.23809523809523808}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"4/21\"],[\"brown\",\"5/21\"],[\"orange\",\"3/7\"],[\"white\",\"1/7\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"orange": 0.93, "brown": 0.01, "white": 0.06, "black": 0}, "focus": {"false": 0.69, "true": 0.31}, "group": {"0": 0.01, "1": 0.97, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:132", "state_id": "known_chance_v2:train:132", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: cyan=1, orange=7, blue=4, black=4, brown=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "brown": "The sampled ball is brown", "cyan": "The sampled ball is cyan", "orange": "The sampled ball is orange", "blue": "The sampled ball is blue"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: cyan, black", "The drawn color is one of: orange, brown", "The drawn color is one of: blue"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.19047619047619047, "brown": 0.23809523809523808, "cyan": 0.047619047619047616, "orange": 0.3333333333333333, "blue": 0.19047619047619047}, "focus": {"false": 0.8095238095238095, "true": 0.19047619047619047}, "group": {"0": 0.23809523809523808, "1": 0.5714285714285714, "2": 0.19047619047619047}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"4/21\"],[\"blue\",\"4/21\"],[\"brown\",\"5/21\"],[\"cyan\",\"1/21\"],[\"orange\",\"1/3\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"orange": 0.95, "brown": 0.01, "cyan": 0.03, "blue": 0, "black": 0.01}, "focus": {"false": 0.81, "true": 0.19}, "group": {"0": 0, "1": 0.99, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:135", "state_id": "known_chance_v2:train:135", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: red=2, blue=7, cyan=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "blue": "The sampled ball is blue", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: red", "The drawn color is one of: blue", "The drawn color is one of: cyan"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.4, "blue": 0.4666666666666667, "red": 0.13333333333333333}, "focus": {"false": 0.5333333333333333, "true": 0.4666666666666667}, "group": {"0": 0.13333333333333333, "1": 0.4666666666666667, "2": 0.4}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"7/15\"],[\"cyan\",\"2/5\"],[\"red\",\"2/15\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0.01, "blue": 0.96, "red": 0.03}, "focus": {"false": 0.61, "true": 0.39}, "group": {"0": 0.01, "1": 0.96, "2": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:134", "state_id": "known_chance_v2:train:134", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: green=1, black=9, pink=1, red=9, cyan=8, orange=1, brown=5, blue=4, purple=9, gray=4, yellow=7, white=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "yellow": "The sampled ball is yellow", "white": "The sampled ball is white", "cyan": "The sampled ball is cyan", "blue": "The sampled ball is blue", "pink": "The sampled ball is pink", "gray": "The sampled ball is gray", "red": "The sampled ball is red", "orange": "The sampled ball is orange", "purple": "The sampled ball is purple", "brown": "The sampled ball is brown", "green": "The sampled ball is green"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a purple ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: green, red, brown, gray", "The drawn color is one of: black, cyan, blue, yellow", "The drawn color is one of: pink, orange, purple, white"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.140625, "yellow": 0.109375, "white": 0.09375, "cyan": 0.125, "blue": 0.0625, "pink": 0.015625, "gray": 0.0625, "red": 0.140625, "orange": 0.015625, "purple": 0.140625, "brown": 0.078125, "green": 0.015625}, "focus": {"false": 0.859375, "true": 0.140625}, "group": {"0": 0.296875, "1": 0.4375, "2": 0.265625}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"9/64\"],[\"blue\",\"1/16\"],[\"brown\",\"5/64\"],[\"cyan\",\"1/8\"],[\"gray\",\"1/16\"],[\"green\",\"1/64\"],[\"orange\",\"1/64\"],[\"pink\",\"1/64\"],[\"purple\",\"9/64\"],[\"red\",\"9/64\"],[\"white\",\"3/32\"],[\"yellow\",\"7/64\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"orange": 0.01, "purple": 0.1, "white": 0.01, "red": 0.16999999999999998, "cyan": 0.05, "pink": 0.01, "gray": 0, "black": 0.58, "brown": 0.01, "yellow": 0.01, "blue": 0.01, "green": 0.04}, "focus": {"false": 0.8200000000000001, "true": 0.18}, "group": {"0": 0.08, "1": 0.89, "2": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:133", "state_id": "known_chance_v2:train:133", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: blue=7, yellow=3, white=9, black=4, brown=5, cyan=3, green=3, pink=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "black": "The sampled ball is black", "blue": "The sampled ball is blue", "green": "The sampled ball is green", "white": "The sampled ball is white", "brown": "The sampled ball is brown", "yellow": "The sampled ball is yellow", "pink": "The sampled ball is pink"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a white ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue, black, green", "The drawn color is one of: yellow, brown, pink", "The drawn color is one of: white, cyan"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.075, "black": 0.1, "blue": 0.175, "green": 0.075, "white": 0.225, "brown": 0.125, "yellow": 0.075, "pink": 0.15}, "focus": {"false": 0.775, "true": 0.225}, "group": {"0": 0.35, "1": 0.35, "2": 0.3}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/10\"],[\"blue\",\"7/40\"],[\"brown\",\"1/8\"],[\"cyan\",\"3/40\"],[\"green\",\"3/40\"],[\"pink\",\"3/20\"],[\"white\",\"9/40\"],[\"yellow\",\"3/40\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"brown": 0, "blue": 0.05, "black": 0, "yellow": 0, "white": 0.95, "cyan": 0, "green": 0, "pink": 0}, "focus": {"false": 0.79, "true": 0.21}, "group": {"0": 0.15, "1": 0.06, "2": 0.79}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:136", "state_id": "known_chance_v2:train:136", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: green=2, orange=7, purple=4, red=9. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"purple": "The sampled ball is purple", "red": "The sampled ball is red", "green": "The sampled ball is green", "orange": "The sampled ball is orange"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a green ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: green, red", "The drawn color is one of: orange", "The drawn color is one of: purple"]}}, "gold": {}, "gold_probs": {"outcome": {"purple": 0.18181818181818182, "red": 0.4090909090909091, "green": 0.09090909090909091, "orange": 0.3181818181818182}, "focus": {"false": 0.9090909090909091, "true": 0.09090909090909091}, "group": {"0": 0.5, "1": 0.3181818181818182, "2": 0.18181818181818182}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"green\",\"1/11\"],[\"orange\",\"7/22\"],[\"purple\",\"2/11\"],[\"red\",\"9/22\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0.07, "red": 0.88, "purple": 0, "orange": 0.05}, "focus": {"false": 0.88, "true": 0.12}, "group": {"0": 0.64, "1": 0.34, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:137", "state_id": "known_chance_v2:train:137", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: gray=4, yellow=6, brown=8, pink=4, green=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"pink": "The sampled ball is pink", "green": "The sampled ball is green", "gray": "The sampled ball is gray", "yellow": "The sampled ball is yellow", "brown": "The sampled ball is brown"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a pink ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: gray, pink", "The drawn color is one of: yellow, green", "The drawn color is one of: brown"]}}, "gold": {}, "gold_probs": {"outcome": {"pink": 0.13793103448275862, "green": 0.2413793103448276, "gray": 0.13793103448275862, "yellow": 0.20689655172413793, "brown": 0.27586206896551724}, "focus": {"false": 0.8620689655172413, "true": 0.13793103448275862}, "group": {"0": 0.27586206896551724, "1": 0.4482758620689655, "2": 0.27586206896551724}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"8/29\"],[\"gray\",\"4/29\"],[\"green\",\"7/29\"],[\"pink\",\"4/29\"],[\"yellow\",\"6/29\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0, "green": 0.01, "gray": 0.05, "yellow": 0.01, "brown": 0.93}, "focus": {"false": 0.85, "true": 0.15}, "group": {"0": 0.02, "1": 0.44, "2": 0.54}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:138", "state_id": "known_chance_v2:train:138", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: cyan=3, green=1, orange=7, blue=2, gray=6, purple=1, pink=9, yellow=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "purple": "The sampled ball is purple", "gray": "The sampled ball is gray", "orange": "The sampled ball is orange", "blue": "The sampled ball is blue", "cyan": "The sampled ball is cyan", "pink": "The sampled ball is pink", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a pink ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: cyan, blue, pink", "The drawn color is one of: green, gray, yellow", "The drawn color is one of: orange, purple"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.03333333333333333, "purple": 0.03333333333333333, "gray": 0.2, "orange": 0.23333333333333334, "blue": 0.06666666666666667, "cyan": 0.1, "pink": 0.3, "yellow": 0.03333333333333333}, "focus": {"false": 0.7, "true": 0.3}, "group": {"0": 0.4666666666666667, "1": 0.26666666666666666, "2": 0.26666666666666666}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"1/15\"],[\"cyan\",\"1/10\"],[\"gray\",\"1/5\"],[\"green\",\"1/30\"],[\"orange\",\"7/30\"],[\"pink\",\"3/10\"],[\"purple\",\"1/30\"],[\"yellow\",\"1/30\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0, "pink": 0.94, "cyan": 0.01, "gray": 0.01, "yellow": 0.01, "blue": 0, "orange": 0.03, "green": 0}, "focus": {"false": 0.79, "true": 0.21}, "group": {"0": 0.79, "1": 0.03, "2": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:140", "state_id": "known_chance_v2:train:140", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: black=8, yellow=2, cyan=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"yellow": "The sampled ball is yellow", "cyan": "The sampled ball is cyan", "black": "The sampled ball is black"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: black", "The drawn color is one of: yellow", "The drawn color is one of: cyan"]}}, "gold": {}, "gold_probs": {"outcome": {"yellow": 0.14285714285714285, "cyan": 0.2857142857142857, "black": 0.5714285714285714}, "focus": {"false": 0.8571428571428572, "true": 0.14285714285714285}, "group": {"0": 0.5714285714285714, "1": 0.14285714285714285, "2": 0.2857142857142857}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"4/7\"],[\"cyan\",\"2/7\"],[\"yellow\",\"1/7\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"yellow": 0.01, "cyan": 0.01, "black": 0.98}, "focus": {"false": 0.83, "true": 0.17}, "group": {"0": 0.99, "1": 0, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:139", "state_id": "known_chance_v2:train:139", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: purple=3, blue=5, orange=1, gray=8, white=6, pink=9, black=9, brown=4, red=7, cyan=3, yellow=2, green=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "purple": "The sampled ball is purple", "cyan": "The sampled ball is cyan", "pink": "The sampled ball is pink", "black": "The sampled ball is black", "orange": "The sampled ball is orange", "red": "The sampled ball is red", "yellow": "The sampled ball is yellow", "brown": "The sampled ball is brown", "green": "The sampled ball is green", "gray": "The sampled ball is gray", "white": "The sampled ball is white"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a gray ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: purple, gray, black, cyan", "The drawn color is one of: blue, white, brown, yellow", "The drawn color is one of: orange, pink, red, green"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.08333333333333333, "purple": 0.05, "cyan": 0.05, "pink": 0.15, "black": 0.15, "orange": 0.016666666666666666, "red": 0.11666666666666667, "yellow": 0.03333333333333333, "brown": 0.06666666666666667, "green": 0.05, "gray": 0.13333333333333333, "white": 0.1}, "focus": {"false": 0.8666666666666667, "true": 0.13333333333333333}, "group": {"0": 0.3833333333333333, "1": 0.2833333333333333, "2": 0.3333333333333333}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"3/20\"],[\"blue\",\"1/12\"],[\"brown\",\"1/15\"],[\"cyan\",\"1/20\"],[\"gray\",\"2/15\"],[\"green\",\"1/20\"],[\"orange\",\"1/60\"],[\"pink\",\"3/20\"],[\"purple\",\"1/20\"],[\"red\",\"7/60\"],[\"white\",\"1/10\"],[\"yellow\",\"1/30\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"blue": 0.01, "red": 0.01, "white": 0.01, "orange": 0.08, "green": 0.01, "gray": 0.32, "black": 0.25, "yellow": 0.01, "pink": 0.26, "purple": 0.03, "brown": 0.01, "cyan": 0}, "focus": {"false": 0.85, "true": 0.15}, "group": {"0": 0.7, "1": 0.07, "2": 0.23}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:143", "state_id": "known_chance_v2:train:143", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: cyan=3, purple=5, red=4, yellow=9, pink=3, white=1, orange=8, green=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "pink": "The sampled ball is pink", "red": "The sampled ball is red", "green": "The sampled ball is green", "purple": "The sampled ball is purple", "white": "The sampled ball is white", "yellow": "The sampled ball is yellow", "orange": "The sampled ball is orange"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a purple ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: cyan, yellow, orange", "The drawn color is one of: purple, pink, green", "The drawn color is one of: red, white"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.08333333333333333, "pink": 0.08333333333333333, "red": 0.1111111111111111, "green": 0.08333333333333333, "purple": 0.1388888888888889, "white": 0.027777777777777776, "yellow": 0.25, "orange": 0.2222222222222222}, "focus": {"false": 0.8611111111111112, "true": 0.1388888888888889}, "group": {"0": 0.5555555555555556, "1": 0.3055555555555556, "2": 0.1388888888888889}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"1/12\"],[\"green\",\"1/12\"],[\"orange\",\"2/9\"],[\"pink\",\"1/12\"],[\"purple\",\"5/36\"],[\"red\",\"1/9\"],[\"white\",\"1/36\"],[\"yellow\",\"1/4\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"red": 0.01, "purple": 0.01, "yellow": 0.88, "orange": 0.04, "white": 0.04, "cyan": 0.02, "green": 0, "pink": 0}, "focus": {"false": 0.83, "true": 0.17}, "group": {"0": 0.98, "1": 0.01, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:141", "state_id": "known_chance_v2:train:141", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: gray=1, pink=9, green=3, black=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"pink": "The sampled ball is pink", "gray": "The sampled ball is gray", "black": "The sampled ball is black", "green": "The sampled ball is green"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a gray ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: gray, black", "The drawn color is one of: pink", "The drawn color is one of: green"]}}, "gold": {}, "gold_probs": {"outcome": {"pink": 0.45, "gray": 0.05, "black": 0.35, "green": 0.15}, "focus": {"false": 0.95, "true": 0.05}, "group": {"0": 0.39999999999999997, "1": 0.45, "2": 0.15}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"7/20\"],[\"gray\",\"1/20\"],[\"green\",\"3/20\"],[\"pink\",\"9/20\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"black": 0.01, "gray": 0.03, "green": 0, "pink": 0.96}, "focus": {"false": 0.94, "true": 0.06}, "group": {"0": 0.02, "1": 0.97, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:145", "state_id": "known_chance_v2:train:145", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: blue=3, pink=8, gray=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "pink": "The sampled ball is pink", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue", "The drawn color is one of: pink", "The drawn color is one of: gray"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.16666666666666666, "pink": 0.4444444444444444, "gray": 0.3888888888888889}, "focus": {"false": 0.8333333333333334, "true": 0.16666666666666666}, "group": {"0": 0.16666666666666666, "1": 0.4444444444444444, "2": 0.3888888888888889}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"1/6\"],[\"gray\",\"7/18\"],[\"pink\",\"4/9\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"gray": 0.01, "pink": 0.92, "blue": 0.07}, "focus": {"false": 0.85, "true": 0.15}, "group": {"0": 0.01, "1": 0.96, "2": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:148", "state_id": "known_chance_v2:train:148", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: white=9, red=6, yellow=3, gray=2, pink=6, blue=8, brown=2, black=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"yellow": "The sampled ball is yellow", "brown": "The sampled ball is brown", "black": "The sampled ball is black", "pink": "The sampled ball is pink", "blue": "The sampled ball is blue", "white": "The sampled ball is white", "red": "The sampled ball is red", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white, gray, brown", "The drawn color is one of: red, pink, black", "The drawn color is one of: yellow, blue"]}}, "gold": {}, "gold_probs": {"outcome": {"yellow": 0.06976744186046512, "brown": 0.046511627906976744, "black": 0.16279069767441862, "pink": 0.13953488372093023, "blue": 0.18604651162790697, "white": 0.20930232558139536, "red": 0.13953488372093023, "gray": 0.046511627906976744}, "focus": {"false": 0.8372093023255813, "true": 0.16279069767441862}, "group": {"0": 0.3023255813953488, "1": 0.4418604651162791, "2": 0.2558139534883721}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"7/43\"],[\"blue\",\"8/43\"],[\"brown\",\"2/43\"],[\"gray\",\"2/43\"],[\"pink\",\"6/43\"],[\"red\",\"6/43\"],[\"white\",\"9/43\"],[\"yellow\",\"3/43\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"red": 0, "brown": 0, "blue": 0.02, "gray": 0, "black": 0, "yellow": 0, "white": 0.98, "pink": 0}, "focus": {"false": 0.83, "true": 0.17}, "group": {"0": 0.32, "1": 0.63, "2": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:147", "state_id": "known_chance_v2:train:147", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: purple=1, red=7, yellow=2, blue=9, green=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"purple": "The sampled ball is purple", "yellow": "The sampled ball is yellow", "red": "The sampled ball is red", "green": "The sampled ball is green", "blue": "The sampled ball is blue"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a red ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: purple, blue", "The drawn color is one of: red, green", "The drawn color is one of: yellow"]}}, "gold": {}, "gold_probs": {"outcome": {"purple": 0.047619047619047616, "yellow": 0.09523809523809523, "red": 0.3333333333333333, "green": 0.09523809523809523, "blue": 0.42857142857142855}, "focus": {"false": 0.6666666666666667, "true": 0.3333333333333333}, "group": {"0": 0.47619047619047616, "1": 0.42857142857142855, "2": 0.09523809523809523}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"3/7\"],[\"green\",\"2/21\"],[\"purple\",\"1/21\"],[\"red\",\"1/3\"],[\"yellow\",\"2/21\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"red": 0.04, "green": 0, "blue": 0.93, "purple": 0.02, "yellow": 0}, "focus": {"false": 0.73, "true": 0.27}, "group": {"0": 0.83, "1": 0.16, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:144", "state_id": "known_chance_v2:train:144", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: brown=2, white=9, green=2, cyan=9, gray=2, black=2, purple=5, blue=1, red=5, yellow=2, orange=5, pink=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"yellow": "The sampled ball is yellow", "black": "The sampled ball is black", "orange": "The sampled ball is orange", "blue": "The sampled ball is blue", "green": "The sampled ball is green", "white": "The sampled ball is white", "gray": "The sampled ball is gray", "purple": "The sampled ball is purple", "red": "The sampled ball is red", "cyan": "The sampled ball is cyan", "brown": "The sampled ball is brown", "pink": "The sampled ball is pink"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a brown ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: brown, cyan, purple, yellow", "The drawn color is one of: white, gray, blue, orange", "The drawn color is one of: green, black, red, pink"]}}, "gold": {}, "gold_probs": {"outcome": {"yellow": 0.043478260869565216, "black": 0.043478260869565216, "orange": 0.10869565217391304, "blue": 0.021739130434782608, "green": 0.043478260869565216, "white": 0.1956521739130435, "gray": 0.043478260869565216, "purple": 0.10869565217391304, "red": 0.10869565217391304, "cyan": 0.1956521739130435, "brown": 0.043478260869565216, "pink": 0.043478260869565216}, "focus": {"false": 0.9565217391304348, "true": 0.043478260869565216}, "group": {"0": 0.391304347826087, "1": 0.3695652173913043, "2": 0.2391304347826087}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/23\"],[\"blue\",\"1/46\"],[\"brown\",\"1/23\"],[\"cyan\",\"9/46\"],[\"gray\",\"1/23\"],[\"green\",\"1/23\"],[\"orange\",\"5/46\"],[\"pink\",\"1/23\"],[\"purple\",\"5/46\"],[\"red\",\"5/46\"],[\"white\",\"9/46\"],[\"yellow\",\"1/23\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0, "cyan": 0.17, "yellow": 0, "white": 0.78, "gray": 0, "purple": 0.01, "black": 0, "green": 0, "blue": 0.02, "red": 0, "brown": 0.02, "orange": 0}, "focus": {"false": 0.9, "true": 0.1}, "group": {"0": 0.62, "1": 0.34, "2": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:146", "state_id": "known_chance_v2:train:146", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: white=1, blue=5, yellow=9, green=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"yellow": "The sampled ball is yellow", "green": "The sampled ball is green", "white": "The sampled ball is white", "blue": "The sampled ball is blue"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white, green", "The drawn color is one of: blue", "The drawn color is one of: yellow"]}}, "gold": {}, "gold_probs": {"outcome": {"yellow": 0.5294117647058824, "green": 0.11764705882352941, "white": 0.058823529411764705, "blue": 0.29411764705882354}, "focus": {"false": 0.7058823529411764, "true": 0.29411764705882354}, "group": {"0": 0.1764705882352941, "1": 0.29411764705882354, "2": 0.5294117647058824}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"5/17\"],[\"green\",\"2/17\"],[\"white\",\"1/17\"],[\"yellow\",\"9/17\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0, "white": 0.07, "blue": 0.02, "yellow": 0.91}, "focus": {"false": 0.81, "true": 0.19}, "group": {"0": 0.01, "1": 0.01, "2": 0.98}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:142", "state_id": "known_chance_v2:train:142", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: yellow=2, pink=4, blue=9, gray=5, black=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"gray": "The sampled ball is gray", "pink": "The sampled ball is pink", "blue": "The sampled ball is blue", "yellow": "The sampled ball is yellow", "black": "The sampled ball is black"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a pink ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: yellow, gray", "The drawn color is one of: pink, black", "The drawn color is one of: blue"]}}, "gold": {}, "gold_probs": {"outcome": {"gray": 0.17857142857142858, "pink": 0.14285714285714285, "blue": 0.32142857142857145, "yellow": 0.07142857142857142, "black": 0.2857142857142857}, "focus": {"false": 0.8571428571428572, "true": 0.14285714285714285}, "group": {"0": 0.25, "1": 0.42857142857142855, "2": 0.32142857142857145}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"2/7\"],[\"blue\",\"9/28\"],[\"gray\",\"5/28\"],[\"pink\",\"1/7\"],[\"yellow\",\"1/14\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"yellow": 0.05, "black": 0, "pink": 0, "blue": 0.95, "gray": 0}, "focus": {"false": 0.83, "true": 0.17}, "group": {"0": 0.01, "1": 0.06, "2": 0.9299999999999999}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:149", "state_id": "known_chance_v2:train:149", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: white=3, yellow=4, purple=2, green=7, red=7, gray=8, blue=4, brown=8, black=1, pink=2, cyan=2, orange=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"white": "The sampled ball is white", "blue": "The sampled ball is blue", "orange": "The sampled ball is orange", "red": "The sampled ball is red", "yellow": "The sampled ball is yellow", "black": "The sampled ball is black", "green": "The sampled ball is green", "purple": "The sampled ball is purple", "gray": "The sampled ball is gray", "pink": "The sampled ball is pink", "brown": "The sampled ball is brown", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a purple ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white, green, blue, pink", "The drawn color is one of: yellow, red, brown, cyan", "The drawn color is one of: purple, gray, black, orange"]}}, "gold": {}, "gold_probs": {"outcome": {"white": 0.058823529411764705, "blue": 0.0784313725490196, "orange": 0.058823529411764705, "red": 0.13725490196078433, "yellow": 0.0784313725490196, "black": 0.0196078431372549, "green": 0.13725490196078433, "purple": 0.0392156862745098, "gray": 0.1568627450980392, "pink": 0.0392156862745098, "brown": 0.1568627450980392, "cyan": 0.0392156862745098}, "focus": {"false": 0.9607843137254902, "true": 0.0392156862745098}, "group": {"0": 0.3137254901960784, "1": 0.4117647058823529, "2": 0.27450980392156865}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/51\"],[\"blue\",\"4/51\"],[\"brown\",\"8/51\"],[\"cyan\",\"2/51\"],[\"gray\",\"8/51\"],[\"green\",\"7/51\"],[\"orange\",\"1/17\"],[\"pink\",\"2/51\"],[\"purple\",\"2/51\"],[\"red\",\"7/51\"],[\"white\",\"1/17\"],[\"yellow\",\"4/51\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"gray": 0.61, "purple": 0, "yellow": 0, "black": 0.05, "cyan": 0.01, "white": 0.06, "blue": 0, "orange": 0.01, "red": 0.02, "green": 0.1, "pink": 0, "brown": 0.14}, "focus": {"false": 0.91, "true": 0.09}, "group": {"0": 0.11, "1": 0.64, "2": 0.25}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:150", "state_id": "known_chance_v2:train:150", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: black=3, yellow=4, blue=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "yellow": "The sampled ball is yellow", "black": "The sampled ball is black"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: black", "The drawn color is one of: yellow", "The drawn color is one of: blue"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.36363636363636365, "yellow": 0.36363636363636365, "black": 0.2727272727272727}, "focus": {"false": 0.7272727272727273, "true": 0.2727272727272727}, "group": {"0": 0.2727272727272727, "1": 0.36363636363636365, "2": 0.36363636363636365}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"3/11\"],[\"blue\",\"4/11\"],[\"yellow\",\"4/11\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"yellow": 0.29, "black": 0.61, "blue": 0.1}, "focus": {"false": 0.75, "true": 0.25}, "group": {"0": 0.1, "1": 0.78, "2": 0.12}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:153", "state_id": "known_chance_v2:train:153", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: orange=7, green=7, black=4, white=5, purple=9, red=5, blue=9, pink=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "orange": "The sampled ball is orange", "pink": "The sampled ball is pink", "purple": "The sampled ball is purple", "white": "The sampled ball is white", "blue": "The sampled ball is blue", "red": "The sampled ball is red", "green": "The sampled ball is green"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a red ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: orange, white, blue", "The drawn color is one of: green, purple, pink", "The drawn color is one of: black, red"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.0851063829787234, "orange": 0.14893617021276595, "pink": 0.02127659574468085, "purple": 0.19148936170212766, "white": 0.10638297872340426, "blue": 0.19148936170212766, "red": 0.10638297872340426, "green": 0.14893617021276595}, "focus": {"false": 0.8936170212765957, "true": 0.10638297872340426}, "group": {"0": 0.44680851063829785, "1": 0.36170212765957444, "2": 0.19148936170212766}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"4/47\"],[\"blue\",\"9/47\"],[\"green\",\"7/47\"],[\"orange\",\"7/47\"],[\"pink\",\"1/47\"],[\"purple\",\"9/47\"],[\"red\",\"5/47\"],[\"white\",\"5/47\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.53, "white": 0.01, "pink": 0.12, "blue": 0.18, "red": 0, "orange": 0.12, "green": 0.03, "black": 0.01}, "focus": {"false": 0.86, "true": 0.14}, "group": {"0": 0.72, "1": 0.26, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:156", "state_id": "known_chance_v2:train:156", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: white=7, purple=6, brown=4, cyan=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"purple": "The sampled ball is purple", "cyan": "The sampled ball is cyan", "white": "The sampled ball is white", "brown": "The sampled ball is brown"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a purple ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white, cyan", "The drawn color is one of: purple", "The drawn color is one of: brown"]}}, "gold": {}, "gold_probs": {"outcome": {"purple": 0.25, "cyan": 0.2916666666666667, "white": 0.2916666666666667, "brown": 0.16666666666666666}, "focus": {"false": 0.75, "true": 0.25}, "group": {"0": 0.5833333333333334, "1": 0.25, "2": 0.16666666666666666}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"1/6\"],[\"cyan\",\"7/24\"],[\"purple\",\"1/4\"],[\"white\",\"7/24\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0.06, "brown": 0.03, "purple": 0.03, "white": 0.88}, "focus": {"false": 0.8, "true": 0.2}, "group": {"0": 0.99, "1": 0.01, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:152", "state_id": "known_chance_v2:train:152", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: green=7, brown=5, blue=2, pink=2, purple=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"pink": "The sampled ball is pink", "green": "The sampled ball is green", "brown": "The sampled ball is brown", "blue": "The sampled ball is blue", "purple": "The sampled ball is purple"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a purple ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: green, pink", "The drawn color is one of: brown, purple", "The drawn color is one of: blue"]}}, "gold": {}, "gold_probs": {"outcome": {"pink": 0.11764705882352941, "green": 0.4117647058823529, "brown": 0.29411764705882354, "blue": 0.11764705882352941, "purple": 0.058823529411764705}, "focus": {"false": 0.9411764705882353, "true": 0.058823529411764705}, "group": {"0": 0.5294117647058824, "1": 0.35294117647058826, "2": 0.11764705882352941}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"2/17\"],[\"brown\",\"5/17\"],[\"green\",\"7/17\"],[\"pink\",\"2/17\"],[\"purple\",\"1/17\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0.99, "blue": 0, "purple": 0.01, "pink": 0, "brown": 0}, "focus": {"false": 0.9299999999999999, "true": 0.07}, "group": {"0": 0.93, "1": 0.06, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:154", "state_id": "known_chance_v2:train:154", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: red=4, green=4, pink=9, orange=8, black=3, white=4, blue=7, brown=6, gray=4, purple=4, yellow=9, cyan=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"purple": "The sampled ball is purple", "orange": "The sampled ball is orange", "green": "The sampled ball is green", "black": "The sampled ball is black", "brown": "The sampled ball is brown", "gray": "The sampled ball is gray", "pink": "The sampled ball is pink", "cyan": "The sampled ball is cyan", "white": "The sampled ball is white", "red": "The sampled ball is red", "yellow": "The sampled ball is yellow", "blue": "The sampled ball is blue"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a white ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: red, orange, blue, purple", "The drawn color is one of: green, black, brown, yellow", "The drawn color is one of: pink, white, gray, cyan"]}}, "gold": {}, "gold_probs": {"outcome": {"purple": 0.05970149253731343, "orange": 0.11940298507462686, "green": 0.05970149253731343, "black": 0.04477611940298507, "brown": 0.08955223880597014, "gray": 0.05970149253731343, "pink": 0.13432835820895522, "cyan": 0.07462686567164178, "white": 0.05970149253731343, "red": 0.05970149253731343, "yellow": 0.13432835820895522, "blue": 0.1044776119402985}, "focus": {"false": 0.9402985074626866, "true": 0.05970149253731343}, "group": {"0": 0.34328358208955223, "1": 0.3283582089552239, "2": 0.3283582089552239}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"3/67\"],[\"blue\",\"7/67\"],[\"brown\",\"6/67\"],[\"cyan\",\"5/67\"],[\"gray\",\"4/67\"],[\"green\",\"4/67\"],[\"orange\",\"8/67\"],[\"pink\",\"9/67\"],[\"purple\",\"4/67\"],[\"red\",\"4/67\"],[\"white\",\"4/67\"],[\"yellow\",\"9/67\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0.04, "green": 0.01, "blue": 0.02, "purple": 0.01, "pink": 0.64, "white": 0, "red": 0.1, "brown": 0.01, "black": 0.01, "orange": 0.02, "gray": 0, "yellow": 0.14}, "focus": {"false": 0.87, "true": 0.13}, "group": {"0": 0.1, "1": 0.16, "2": 0.74}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:155", "state_id": "known_chance_v2:train:155", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: blue=1, purple=8, black=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "black": "The sampled ball is black", "purple": "The sampled ball is purple"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue", "The drawn color is one of: purple", "The drawn color is one of: black"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.058823529411764705, "black": 0.47058823529411764, "purple": 0.47058823529411764}, "focus": {"false": 0.5294117647058824, "true": 0.47058823529411764}, "group": {"0": 0.058823529411764705, "1": 0.47058823529411764, "2": 0.47058823529411764}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"8/17\"],[\"blue\",\"1/17\"],[\"purple\",\"8/17\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"blue": 0.25, "black": 0.11, "purple": 0.64}, "focus": {"false": 0.72, "true": 0.28}, "group": {"0": 0.02, "1": 0.82, "2": 0.16}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:157", "state_id": "known_chance_v2:train:157", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: black=3, yellow=6, blue=3, pink=1, green=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "blue": "The sampled ball is blue", "green": "The sampled ball is green", "yellow": "The sampled ball is yellow", "pink": "The sampled ball is pink"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: black, pink", "The drawn color is one of: yellow, green", "The drawn color is one of: blue"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.17647058823529413, "blue": 0.17647058823529413, "green": 0.23529411764705882, "yellow": 0.35294117647058826, "pink": 0.058823529411764705}, "focus": {"false": 0.6470588235294117, "true": 0.35294117647058826}, "group": {"0": 0.23529411764705882, "1": 0.5882352941176471, "2": 0.17647058823529413}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"3/17\"],[\"blue\",\"3/17\"],[\"green\",\"4/17\"],[\"pink\",\"1/17\"],[\"yellow\",\"6/17\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.01, "black": 0.05, "green": 0.01, "yellow": 0.93, "blue": 0}, "focus": {"false": 0.69, "true": 0.31}, "group": {"0": 0.01, "1": 0.99, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:158", "state_id": "known_chance_v2:train:158", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: purple=3, pink=7, cyan=9, orange=4, red=9, yellow=8, black=9, gray=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"pink": "The sampled ball is pink", "gray": "The sampled ball is gray", "orange": "The sampled ball is orange", "yellow": "The sampled ball is yellow", "cyan": "The sampled ball is cyan", "black": "The sampled ball is black", "red": "The sampled ball is red", "purple": "The sampled ball is purple"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: purple, orange, black", "The drawn color is one of: pink, red, gray", "The drawn color is one of: cyan, yellow"]}}, "gold": {}, "gold_probs": {"outcome": {"pink": 0.1346153846153846, "gray": 0.057692307692307696, "orange": 0.07692307692307693, "yellow": 0.15384615384615385, "cyan": 0.17307692307692307, "black": 0.17307692307692307, "red": 0.17307692307692307, "purple": 0.057692307692307696}, "focus": {"false": 0.8269230769230769, "true": 0.17307692307692307}, "group": {"0": 0.3076923076923077, "1": 0.36538461538461536, "2": 0.3269230769230769}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"9/52\"],[\"cyan\",\"9/52\"],[\"gray\",\"3/52\"],[\"orange\",\"1/13\"],[\"pink\",\"7/52\"],[\"purple\",\"3/52\"],[\"red\",\"9/52\"],[\"yellow\",\"2/13\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.08, "cyan": 0.62, "red": 0.11, "orange": 0.01, "gray": 0.01, "black": 0.11, "pink": 0.03, "yellow": 0.03}, "focus": {"false": 0.8, "true": 0.2}, "group": {"0": 0.09, "1": 0.18, "2": 0.73}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:151", "state_id": "known_chance_v2:train:151", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: cyan=4, red=2, gray=7, blue=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"gray": "The sampled ball is gray", "red": "The sampled ball is red", "blue": "The sampled ball is blue", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: cyan, blue", "The drawn color is one of: red", "The drawn color is one of: gray"]}}, "gold": {}, "gold_probs": {"outcome": {"gray": 0.4666666666666667, "red": 0.13333333333333333, "blue": 0.13333333333333333, "cyan": 0.26666666666666666}, "focus": {"false": 0.8666666666666667, "true": 0.13333333333333333}, "group": {"0": 0.4, "1": 0.13333333333333333, "2": 0.4666666666666667}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"2/15\"],[\"cyan\",\"4/15\"],[\"gray\",\"7/15\"],[\"red\",\"2/15\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"red": 0, "cyan": 0.04, "blue": 0, "gray": 0.96}, "focus": {"false": 0.86, "true": 0.14}, "group": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:166", "state_id": "known_chance_v2:train:166", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: black=3, gray=6, pink=3, orange=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "orange": "The sampled ball is orange", "pink": "The sampled ball is pink", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: black, orange", "The drawn color is one of: gray", "The drawn color is one of: pink"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.16666666666666666, "orange": 0.3333333333333333, "pink": 0.16666666666666666, "gray": 0.3333333333333333}, "focus": {"false": 0.8333333333333334, "true": 0.16666666666666666}, "group": {"0": 0.5, "1": 0.3333333333333333, "2": 0.16666666666666666}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/6\"],[\"gray\",\"1/3\"],[\"orange\",\"1/3\"],[\"pink\",\"1/6\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"orange": 0.07, "pink": 0.02, "gray": 0.57, "black": 0.34}, "focus": {"false": 0.8200000000000001, "true": 0.18}, "group": {"0": 0.36, "1": 0.63, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:165", "state_id": "known_chance_v2:train:165", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: cyan=9, green=7, pink=9. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "cyan": "The sampled ball is cyan", "pink": "The sampled ball is pink"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a pink ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: cyan", "The drawn color is one of: green", "The drawn color is one of: pink"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.28, "cyan": 0.36, "pink": 0.36}, "focus": {"false": 0.64, "true": 0.36}, "group": {"0": 0.36, "1": 0.28, "2": 0.36}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"9/25\"],[\"green\",\"7/25\"],[\"pink\",\"9/25\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0.07, "pink": 0.05, "cyan": 0.88}, "focus": {"false": 0.63, "true": 0.37}, "group": {"0": 0.79, "1": 0.07, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:161", "state_id": "known_chance_v2:train:161", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: pink=4, green=3, yellow=1, white=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "pink": "The sampled ball is pink", "yellow": "The sampled ball is yellow", "white": "The sampled ball is white"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a white ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: pink, white", "The drawn color is one of: green", "The drawn color is one of: yellow"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.21428571428571427, "pink": 0.2857142857142857, "yellow": 0.07142857142857142, "white": 0.42857142857142855}, "focus": {"false": 0.5714285714285714, "true": 0.42857142857142855}, "group": {"0": 0.7142857142857142, "1": 0.21428571428571427, "2": 0.07142857142857142}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"green\",\"3/14\"],[\"pink\",\"2/7\"],[\"white\",\"3/7\"],[\"yellow\",\"1/14\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"yellow": 0.01, "green": 0.01, "pink": 0.07, "white": 0.91}, "focus": {"false": 0.69, "true": 0.31}, "group": {"0": 0.99, "1": 0, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:159", "state_id": "known_chance_v2:train:159", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: yellow=3, red=6, purple=6, green=1, gray=9, white=1, cyan=5, brown=2, pink=9, black=9, blue=8, orange=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "red": "The sampled ball is red", "gray": "The sampled ball is gray", "yellow": "The sampled ball is yellow", "orange": "The sampled ball is orange", "pink": "The sampled ball is pink", "white": "The sampled ball is white", "green": "The sampled ball is green", "black": "The sampled ball is black", "purple": "The sampled ball is purple", "brown": "The sampled ball is brown", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: yellow, green, cyan, black", "The drawn color is one of: red, gray, brown, blue", "The drawn color is one of: purple, white, pink, orange"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.13333333333333333, "red": 0.1, "gray": 0.15, "yellow": 0.05, "orange": 0.016666666666666666, "pink": 0.15, "white": 0.016666666666666666, "green": 0.016666666666666666, "black": 0.15, "purple": 0.1, "brown": 0.03333333333333333, "cyan": 0.08333333333333333}, "focus": {"false": 0.85, "true": 0.15}, "group": {"0": 0.3, "1": 0.4166666666666667, "2": 0.2833333333333333}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"3/20\"],[\"blue\",\"2/15\"],[\"brown\",\"1/30\"],[\"cyan\",\"1/12\"],[\"gray\",\"3/20\"],[\"green\",\"1/60\"],[\"orange\",\"1/60\"],[\"pink\",\"3/20\"],[\"purple\",\"1/10\"],[\"red\",\"1/10\"],[\"white\",\"1/60\"],[\"yellow\",\"1/20\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.18, "black": 0.11, "cyan": 0.01, "brown": 0, "purple": 0.02, "gray": 0.5, "green": 0.01, "blue": 0.03, "yellow": 0.09, "red": 0.03, "white": 0.01, "orange": 0.01}, "focus": {"false": 0.78, "true": 0.22}, "group": {"0": 0.15, "1": 0.71, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:164", "state_id": "known_chance_v2:train:164", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: yellow=9, blue=8, green=7, white=3, cyan=3, purple=5, orange=2, pink=8, black=5, gray=9, brown=8, red=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"yellow": "The sampled ball is yellow", "pink": "The sampled ball is pink", "blue": "The sampled ball is blue", "purple": "The sampled ball is purple", "gray": "The sampled ball is gray", "black": "The sampled ball is black", "cyan": "The sampled ball is cyan", "red": "The sampled ball is red", "brown": "The sampled ball is brown", "green": "The sampled ball is green", "orange": "The sampled ball is orange", "white": "The sampled ball is white"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: yellow, white, orange, gray", "The drawn color is one of: blue, cyan, pink, brown", "The drawn color is one of: green, purple, black, red"]}}, "gold": {}, "gold_probs": {"outcome": {"yellow": 0.12857142857142856, "pink": 0.11428571428571428, "blue": 0.11428571428571428, "purple": 0.07142857142857142, "gray": 0.12857142857142856, "black": 0.07142857142857142, "cyan": 0.04285714285714286, "red": 0.04285714285714286, "brown": 0.11428571428571428, "green": 0.1, "orange": 0.02857142857142857, "white": 0.04285714285714286}, "focus": {"false": 0.9285714285714286, "true": 0.07142857142857142}, "group": {"0": 0.32857142857142857, "1": 0.3857142857142857, "2": 0.2857142857142857}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/14\"],[\"blue\",\"4/35\"],[\"brown\",\"4/35\"],[\"cyan\",\"3/70\"],[\"gray\",\"9/70\"],[\"green\",\"1/10\"],[\"orange\",\"1/35\"],[\"pink\",\"4/35\"],[\"purple\",\"1/14\"],[\"red\",\"3/70\"],[\"white\",\"3/70\"],[\"yellow\",\"9/70\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0, "cyan": 0.01, "white": 0.01, "pink": 0.01, "gray": 0.13, "blue": 0.04, "yellow": 0.71, "brown": 0.01, "red": 0.02, "orange": 0.04, "green": 0.02, "black": 0}, "focus": {"false": 0.9, "true": 0.1}, "group": {"0": 0.47, "1": 0.47, "2": 0.06}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:162", "state_id": "known_chance_v2:train:162", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: orange=4, gray=5, green=9, pink=1, yellow=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"pink": "The sampled ball is pink", "orange": "The sampled ball is orange", "gray": "The sampled ball is gray", "yellow": "The sampled ball is yellow", "green": "The sampled ball is green"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: orange, pink", "The drawn color is one of: gray, yellow", "The drawn color is one of: green"]}}, "gold": {}, "gold_probs": {"outcome": {"pink": 0.037037037037037035, "orange": 0.14814814814814814, "gray": 0.18518518518518517, "yellow": 0.2962962962962963, "green": 0.3333333333333333}, "focus": {"false": 0.7037037037037037, "true": 0.2962962962962963}, "group": {"0": 0.18518518518518517, "1": 0.48148148148148145, "2": 0.3333333333333333}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"gray\",\"5/27\"],[\"green\",\"1/3\"],[\"orange\",\"4/27\"],[\"pink\",\"1/27\"],[\"yellow\",\"8/27\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"gray": 0.01, "pink": 0.01, "orange": 0.01, "yellow": 0, "green": 0.97}, "focus": {"false": 0.83, "true": 0.17}, "group": {"0": 0, "1": 0.08, "2": 0.92}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:163", "state_id": "known_chance_v2:train:163", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: orange=3, gray=4, yellow=4, pink=3, red=2, blue=5, white=1, purple=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"orange": "The sampled ball is orange", "yellow": "The sampled ball is yellow", "pink": "The sampled ball is pink", "gray": "The sampled ball is gray", "red": "The sampled ball is red", "blue": "The sampled ball is blue", "white": "The sampled ball is white", "purple": "The sampled ball is purple"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: orange, pink, white", "The drawn color is one of: gray, red, purple", "The drawn color is one of: yellow, blue"]}}, "gold": {}, "gold_probs": {"outcome": {"orange": 0.10344827586206896, "yellow": 0.13793103448275862, "pink": 0.10344827586206896, "gray": 0.13793103448275862, "red": 0.06896551724137931, "blue": 0.1724137931034483, "white": 0.034482758620689655, "purple": 0.2413793103448276}, "focus": {"false": 0.8275862068965517, "true": 0.1724137931034483}, "group": {"0": 0.24137931034482757, "1": 0.4482758620689655, "2": 0.3103448275862069}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"5/29\"],[\"gray\",\"4/29\"],[\"orange\",\"3/29\"],[\"pink\",\"3/29\"],[\"purple\",\"7/29\"],[\"red\",\"2/29\"],[\"white\",\"1/29\"],[\"yellow\",\"4/29\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.93, "pink": 0, "orange": 0.02, "blue": 0.02, "red": 0, "white": 0.02, "yellow": 0, "gray": 0.01}, "focus": {"false": 0.8200000000000001, "true": 0.18}, "group": {"0": 0.02, "1": 0.6900000000000001, "2": 0.29}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:160", "state_id": "known_chance_v2:train:160", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: yellow=5, cyan=1, orange=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"yellow": "The sampled ball is yellow", "orange": "The sampled ball is orange", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: yellow", "The drawn color is one of: cyan", "The drawn color is one of: orange"]}}, "gold": {}, "gold_probs": {"outcome": {"yellow": 0.45454545454545453, "orange": 0.45454545454545453, "cyan": 0.09090909090909091}, "focus": {"false": 0.5454545454545454, "true": 0.45454545454545453}, "group": {"0": 0.45454545454545453, "1": 0.09090909090909091, "2": 0.45454545454545453}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"1/11\"],[\"orange\",\"5/11\"],[\"yellow\",\"5/11\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"orange": 0.04, "cyan": 0.04, "yellow": 0.92}, "focus": {"false": 0.63, "true": 0.37}, "group": {"0": 0.87, "1": 0.04, "2": 0.09}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:167", "state_id": "known_chance_v2:train:167", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: yellow=9, gray=8, pink=7, black=6, brown=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"brown": "The sampled ball is brown", "gray": "The sampled ball is gray", "yellow": "The sampled ball is yellow", "pink": "The sampled ball is pink", "black": "The sampled ball is black"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a pink ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: yellow, black", "The drawn color is one of: gray, brown", "The drawn color is one of: pink"]}}, "gold": {}, "gold_probs": {"outcome": {"brown": 0.09090909090909091, "gray": 0.24242424242424243, "yellow": 0.2727272727272727, "pink": 0.21212121212121213, "black": 0.18181818181818182}, "focus": {"false": 0.7878787878787878, "true": 0.21212121212121213}, "group": {"0": 0.45454545454545453, "1": 0.33333333333333337, "2": 0.21212121212121213}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"2/11\"],[\"brown\",\"1/11\"],[\"gray\",\"8/33\"],[\"pink\",\"7/33\"],[\"yellow\",\"3/11\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"brown": 0.04, "pink": 0, "gray": 0.01, "yellow": 0.95, "black": 0}, "focus": {"false": 0.81, "true": 0.19}, "group": {"0": 0.96, "1": 0.02, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:168", "state_id": "known_chance_v2:train:168", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: blue=3, gray=8, orange=3, red=4, purple=3, brown=7, yellow=7, pink=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"red": "The sampled ball is red", "purple": "The sampled ball is purple", "blue": "The sampled ball is blue", "orange": "The sampled ball is orange", "pink": "The sampled ball is pink", "brown": "The sampled ball is brown", "yellow": "The sampled ball is yellow", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a pink ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue, red, yellow", "The drawn color is one of: gray, purple, pink", "The drawn color is one of: orange, brown"]}}, "gold": {}, "gold_probs": {"outcome": {"red": 0.10526315789473684, "purple": 0.07894736842105263, "blue": 0.07894736842105263, "orange": 0.07894736842105263, "pink": 0.07894736842105263, "brown": 0.18421052631578946, "yellow": 0.18421052631578946, "gray": 0.21052631578947367}, "focus": {"false": 0.9210526315789473, "true": 0.07894736842105263}, "group": {"0": 0.3684210526315789, "1": 0.3684210526315789, "2": 0.2631578947368421}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"3/38\"],[\"brown\",\"7/38\"],[\"gray\",\"4/19\"],[\"orange\",\"3/38\"],[\"pink\",\"3/38\"],[\"purple\",\"3/38\"],[\"red\",\"2/19\"],[\"yellow\",\"7/38\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0, "purple": 0, "yellow": 0, "red": 0, "gray": 0.98, "blue": 0.02, "orange": 0, "brown": 0}, "focus": {"false": 0.88, "true": 0.12}, "group": {"0": 0.15, "1": 0.76, "2": 0.09}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:169", "state_id": "known_chance_v2:train:169", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: yellow=3, brown=8, white=4, green=3, purple=7, orange=3, black=6, red=6, gray=2, cyan=5, blue=1, pink=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"pink": "The sampled ball is pink", "purple": "The sampled ball is purple", "white": "The sampled ball is white", "green": "The sampled ball is green", "cyan": "The sampled ball is cyan", "brown": "The sampled ball is brown", "orange": "The sampled ball is orange", "yellow": "The sampled ball is yellow", "gray": "The sampled ball is gray", "black": "The sampled ball is black", "red": "The sampled ball is red", "blue": "The sampled ball is blue"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a purple ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: yellow, green, black, cyan", "The drawn color is one of: brown, purple, red, blue", "The drawn color is one of: white, orange, gray, pink"]}}, "gold": {}, "gold_probs": {"outcome": {"pink": 0.058823529411764705, "purple": 0.13725490196078433, "white": 0.0784313725490196, "green": 0.058823529411764705, "cyan": 0.09803921568627451, "brown": 0.1568627450980392, "orange": 0.058823529411764705, "yellow": 0.058823529411764705, "gray": 0.0392156862745098, "black": 0.11764705882352941, "red": 0.11764705882352941, "blue": 0.0196078431372549}, "focus": {"false": 0.8627450980392157, "true": 0.13725490196078433}, "group": {"0": 0.3333333333333333, "1": 0.43137254901960786, "2": 0.23529411764705882}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"2/17\"],[\"blue\",\"1/51\"],[\"brown\",\"8/51\"],[\"cyan\",\"5/51\"],[\"gray\",\"2/51\"],[\"green\",\"1/17\"],[\"orange\",\"1/17\"],[\"pink\",\"1/17\"],[\"purple\",\"7/51\"],[\"red\",\"2/17\"],[\"white\",\"4/51\"],[\"yellow\",\"1/17\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0, "pink": 0, "yellow": 0.01, "red": 0, "cyan": 0, "green": 0, "black": 0, "brown": 0.91, "orange": 0, "purple": 0.01, "gray": 0, "blue": 0.06999999999999999}, "focus": {"false": 0.84, "true": 0.16}, "group": {"0": 0.06, "1": 0.92, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:173", "state_id": "known_chance_v2:train:173", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: brown=1, yellow=4, purple=7, red=4, white=9, pink=6, orange=9, cyan=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"red": "The sampled ball is red", "brown": "The sampled ball is brown", "white": "The sampled ball is white", "yellow": "The sampled ball is yellow", "orange": "The sampled ball is orange", "cyan": "The sampled ball is cyan", "pink": "The sampled ball is pink", "purple": "The sampled ball is purple"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a orange ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: brown, red, orange", "The drawn color is one of: yellow, white, cyan", "The drawn color is one of: purple, pink"]}}, "gold": {}, "gold_probs": {"outcome": {"red": 0.09523809523809523, "brown": 0.023809523809523808, "white": 0.21428571428571427, "yellow": 0.09523809523809523, "orange": 0.21428571428571427, "cyan": 0.047619047619047616, "pink": 0.14285714285714285, "purple": 0.16666666666666666}, "focus": {"false": 0.7857142857142857, "true": 0.21428571428571427}, "group": {"0": 0.3333333333333333, "1": 0.3571428571428571, "2": 0.30952380952380953}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"1/42\"],[\"cyan\",\"1/21\"],[\"orange\",\"3/14\"],[\"pink\",\"1/7\"],[\"purple\",\"1/6\"],[\"red\",\"2/21\"],[\"white\",\"3/14\"],[\"yellow\",\"2/21\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"orange": 0.12, "purple": 0.03, "white": 0.6799999999999999, "red": 0.01, "cyan": 0.01, "pink": 0.01, "yellow": 0.01, "brown": 0.13}, "focus": {"false": 0.83, "true": 0.17}, "group": {"0": 0.13, "1": 0.68, "2": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:170", "state_id": "known_chance_v2:train:170", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: green=2, pink=8, purple=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "purple": "The sampled ball is purple", "pink": "The sampled ball is pink"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a green ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: green", "The drawn color is one of: pink", "The drawn color is one of: purple"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.125, "purple": 0.375, "pink": 0.5}, "focus": {"false": 0.875, "true": 0.125}, "group": {"0": 0.125, "1": 0.5, "2": 0.375}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"green\",\"1/8\"],[\"pink\",\"1/2\"],[\"purple\",\"3/8\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.01, "pink": 0.92, "green": 0.07}, "focus": {"false": 0.86, "true": 0.14}, "group": {"0": 0.01, "1": 0.98, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:174", "state_id": "known_chance_v2:train:174", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: white=3, pink=4, green=9, blue=3, cyan=8, red=1, gray=9, purple=5, black=8, yellow=7, orange=8, brown=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "brown": "The sampled ball is brown", "red": "The sampled ball is red", "purple": "The sampled ball is purple", "blue": "The sampled ball is blue", "white": "The sampled ball is white", "gray": "The sampled ball is gray", "pink": "The sampled ball is pink", "orange": "The sampled ball is orange", "yellow": "The sampled ball is yellow", "cyan": "The sampled ball is cyan", "black": "The sampled ball is black"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white, blue, gray, yellow", "The drawn color is one of: pink, cyan, purple, orange", "The drawn color is one of: green, red, black, brown"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.12857142857142856, "brown": 0.07142857142857142, "red": 0.014285714285714285, "purple": 0.07142857142857142, "blue": 0.04285714285714286, "white": 0.04285714285714286, "gray": 0.12857142857142856, "pink": 0.05714285714285714, "orange": 0.11428571428571428, "yellow": 0.1, "cyan": 0.11428571428571428, "black": 0.11428571428571428}, "focus": {"false": 0.9571428571428572, "true": 0.04285714285714286}, "group": {"0": 0.3142857142857143, "1": 0.35714285714285715, "2": 0.32857142857142857}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"4/35\"],[\"blue\",\"3/70\"],[\"brown\",\"1/14\"],[\"cyan\",\"4/35\"],[\"gray\",\"9/70\"],[\"green\",\"9/70\"],[\"orange\",\"4/35\"],[\"pink\",\"2/35\"],[\"purple\",\"1/14\"],[\"red\",\"1/70\"],[\"white\",\"3/70\"],[\"yellow\",\"1/10\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0.06, "cyan": 0.02, "black": 0.02, "red": 0.11, "green": 0.53, "orange": 0.01, "yellow": 0.01, "blue": 0.01, "brown": 0.01, "pink": 0.01, "purple": 0.01, "gray": 0.2}, "focus": {"false": 0.89, "true": 0.11}, "group": {"0": 0.2, "1": 0.16, "2": 0.64}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:171", "state_id": "known_chance_v2:train:171", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: pink=4, purple=2, blue=5, cyan=9. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"purple": "The sampled ball is purple", "blue": "The sampled ball is blue", "pink": "The sampled ball is pink", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a pink ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: pink, cyan", "The drawn color is one of: purple", "The drawn color is one of: blue"]}}, "gold": {}, "gold_probs": {"outcome": {"purple": 0.1, "blue": 0.25, "pink": 0.2, "cyan": 0.45}, "focus": {"false": 0.8, "true": 0.2}, "group": {"0": 0.65, "1": 0.1, "2": 0.25}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"1/4\"],[\"cyan\",\"9/20\"],[\"pink\",\"1/5\"],[\"purple\",\"1/10\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.07, "blue": 0.01, "cyan": 0.91, "purple": 0.01}, "focus": {"false": 0.8200000000000001, "true": 0.18}, "group": {"0": 0.92, "1": 0.01, "2": 0.07}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:172", "state_id": "known_chance_v2:train:172", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: brown=5, cyan=9, orange=5, blue=7, white=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"white": "The sampled ball is white", "brown": "The sampled ball is brown", "orange": "The sampled ball is orange", "blue": "The sampled ball is blue", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a white ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: brown, blue", "The drawn color is one of: cyan, white", "The drawn color is one of: orange"]}}, "gold": {}, "gold_probs": {"outcome": {"white": 0.23529411764705882, "brown": 0.14705882352941177, "orange": 0.14705882352941177, "blue": 0.20588235294117646, "cyan": 0.2647058823529412}, "focus": {"false": 0.7647058823529411, "true": 0.23529411764705882}, "group": {"0": 0.3529411764705882, "1": 0.5, "2": 0.14705882352941177}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"7/34\"],[\"brown\",\"5/34\"],[\"cyan\",\"9/34\"],[\"orange\",\"5/34\"],[\"white\",\"4/17\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"blue": 0.01, "white": 0.01, "cyan": 0.92, "brown": 0.06, "orange": 0}, "focus": {"false": 0.8200000000000001, "true": 0.18}, "group": {"0": 0.01, "1": 0.99, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:175", "state_id": "known_chance_v2:train:175", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: blue=2, cyan=4, red=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"red": "The sampled ball is red", "blue": "The sampled ball is blue", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue", "The drawn color is one of: cyan", "The drawn color is one of: red"]}}, "gold": {}, "gold_probs": {"outcome": {"red": 0.5384615384615384, "blue": 0.15384615384615385, "cyan": 0.3076923076923077}, "focus": {"false": 0.6923076923076923, "true": 0.3076923076923077}, "group": {"0": 0.15384615384615385, "1": 0.3076923076923077, "2": 0.5384615384615384}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"2/13\"],[\"cyan\",\"4/13\"],[\"red\",\"7/13\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0.02, "blue": 0.03, "red": 0.95}, "focus": {"false": 0.72, "true": 0.28}, "group": {"0": 0.01, "1": 0.02, "2": 0.97}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:177", "state_id": "known_chance_v2:train:177", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: brown=9, gray=3, purple=4, blue=3, orange=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"brown": "The sampled ball is brown", "purple": "The sampled ball is purple", "gray": "The sampled ball is gray", "blue": "The sampled ball is blue", "orange": "The sampled ball is orange"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a brown ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: brown, blue", "The drawn color is one of: gray, orange", "The drawn color is one of: purple"]}}, "gold": {}, "gold_probs": {"outcome": {"brown": 0.375, "purple": 0.16666666666666666, "gray": 0.125, "blue": 0.125, "orange": 0.20833333333333334}, "focus": {"false": 0.625, "true": 0.375}, "group": {"0": 0.5, "1": 0.33333333333333337, "2": 0.16666666666666666}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"1/8\"],[\"brown\",\"3/8\"],[\"gray\",\"1/8\"],[\"orange\",\"5/24\"],[\"purple\",\"1/6\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"gray": 0, "purple": 0, "blue": 0, "brown": 0.99, "orange": 0.01}, "focus": {"false": 0.6799999999999999, "true": 0.32}, "group": {"0": 0.95, "1": 0.02, "2": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:176", "state_id": "known_chance_v2:train:176", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: white=5, orange=3, red=6, black=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"white": "The sampled ball is white", "black": "The sampled ball is black", "orange": "The sampled ball is orange", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white, black", "The drawn color is one of: orange", "The drawn color is one of: red"]}}, "gold": {}, "gold_probs": {"outcome": {"white": 0.2777777777777778, "black": 0.2222222222222222, "orange": 0.16666666666666666, "red": 0.3333333333333333}, "focus": {"false": 0.7777777777777778, "true": 0.2222222222222222}, "group": {"0": 0.5, "1": 0.16666666666666666, "2": 0.3333333333333333}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"2/9\"],[\"orange\",\"1/6\"],[\"red\",\"1/3\"],[\"white\",\"5/18\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0.32, "orange": 0.01, "black": 0.01, "red": 0.66}, "focus": {"false": 0.81, "true": 0.19}, "group": {"0": 0.12, "1": 0.01, "2": 0.87}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:178", "state_id": "known_chance_v2:train:178", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: gray=4, brown=8, green=4, red=4, cyan=1, black=7, blue=1, pink=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "gray": "The sampled ball is gray", "blue": "The sampled ball is blue", "pink": "The sampled ball is pink", "red": "The sampled ball is red", "green": "The sampled ball is green", "brown": "The sampled ball is brown", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: gray, red, blue", "The drawn color is one of: brown, cyan, pink", "The drawn color is one of: green, black"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.22580645161290322, "gray": 0.12903225806451613, "blue": 0.03225806451612903, "pink": 0.06451612903225806, "red": 0.12903225806451613, "green": 0.12903225806451613, "brown": 0.25806451612903225, "cyan": 0.03225806451612903}, "focus": {"false": 0.967741935483871, "true": 0.03225806451612903}, "group": {"0": 0.29032258064516125, "1": 0.3548387096774194, "2": 0.3548387096774194}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"7/31\"],[\"blue\",\"1/31\"],[\"brown\",\"8/31\"],[\"cyan\",\"1/31\"],[\"gray\",\"4/31\"],[\"green\",\"4/31\"],[\"pink\",\"2/31\"],[\"red\",\"4/31\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0, "green": 0, "blue": 0.01, "red": 0, "cyan": 0.01, "black": 0.03, "brown": 0.94, "gray": 0.01}, "focus": {"false": 0.95, "true": 0.05}, "group": {"0": 0.05, "1": 0.53, "2": 0.42}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:181", "state_id": "known_chance_v2:train:181", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: black=5, brown=3, purple=4, red=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"purple": "The sampled ball is purple", "black": "The sampled ball is black", "brown": "The sampled ball is brown", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a brown ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: black, red", "The drawn color is one of: brown", "The drawn color is one of: purple"]}}, "gold": {}, "gold_probs": {"outcome": {"purple": 0.2222222222222222, "black": 0.2777777777777778, "brown": 0.16666666666666666, "red": 0.3333333333333333}, "focus": {"false": 0.8333333333333334, "true": 0.16666666666666666}, "group": {"0": 0.6111111111111112, "1": 0.16666666666666666, "2": 0.2222222222222222}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"5/18\"],[\"brown\",\"1/6\"],[\"purple\",\"2/9\"],[\"red\",\"1/3\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"red": 0.74, "brown": 0.01, "black": 0.24, "purple": 0.01}, "focus": {"false": 0.8, "true": 0.2}, "group": {"0": 0.98, "1": 0.01, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:180", "state_id": "known_chance_v2:train:180", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: gray=4, black=3, blue=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"gray": "The sampled ball is gray", "black": "The sampled ball is black", "blue": "The sampled ball is blue"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: gray", "The drawn color is one of: black", "The drawn color is one of: blue"]}}, "gold": {}, "gold_probs": {"outcome": {"gray": 0.4444444444444444, "black": 0.3333333333333333, "blue": 0.2222222222222222}, "focus": {"false": 0.6666666666666667, "true": 0.3333333333333333}, "group": {"0": 0.4444444444444444, "1": 0.3333333333333333, "2": 0.2222222222222222}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/3\"],[\"blue\",\"2/9\"],[\"gray\",\"4/9\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"blue": 0.01, "gray": 0.98, "black": 0.01}, "focus": {"false": 0.69, "true": 0.31}, "group": {"0": 0.97, "1": 0.02, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:182", "state_id": "known_chance_v2:train:182", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: orange=7, purple=9, white=7, red=2, cyan=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"orange": "The sampled ball is orange", "white": "The sampled ball is white", "red": "The sampled ball is red", "purple": "The sampled ball is purple", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a red ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: orange, red", "The drawn color is one of: purple, cyan", "The drawn color is one of: white"]}}, "gold": {}, "gold_probs": {"outcome": {"orange": 0.25, "white": 0.25, "red": 0.07142857142857142, "purple": 0.32142857142857145, "cyan": 0.10714285714285714}, "focus": {"false": 0.9285714285714286, "true": 0.07142857142857142}, "group": {"0": 0.3214285714285714, "1": 0.4285714285714286, "2": 0.25}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"3/28\"],[\"orange\",\"1/4\"],[\"purple\",\"9/28\"],[\"red\",\"1/14\"],[\"white\",\"1/4\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.89, "red": 0.01, "orange": 0.09, "white": 0.01, "cyan": 0}, "focus": {"false": 0.9, "true": 0.1}, "group": {"0": 0.04, "1": 0.92, "2": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:183", "state_id": "known_chance_v2:train:183", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: purple=5, black=6, gray=7, yellow=5, green=9, cyan=9, pink=6, blue=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"purple": "The sampled ball is purple", "black": "The sampled ball is black", "yellow": "The sampled ball is yellow", "green": "The sampled ball is green", "blue": "The sampled ball is blue", "cyan": "The sampled ball is cyan", "gray": "The sampled ball is gray", "pink": "The sampled ball is pink"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: purple, yellow, pink", "The drawn color is one of: black, green, blue", "The drawn color is one of: gray, cyan"]}}, "gold": {}, "gold_probs": {"outcome": {"purple": 0.09259259259259259, "black": 0.1111111111111111, "yellow": 0.09259259259259259, "green": 0.16666666666666666, "blue": 0.12962962962962962, "cyan": 0.16666666666666666, "gray": 0.12962962962962962, "pink": 0.1111111111111111}, "focus": {"false": 0.8703703703703703, "true": 0.12962962962962962}, "group": {"0": 0.2962962962962963, "1": 0.4074074074074074, "2": 0.2962962962962963}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/9\"],[\"blue\",\"7/54\"],[\"cyan\",\"1/6\"],[\"gray\",\"7/54\"],[\"green\",\"1/6\"],[\"pink\",\"1/9\"],[\"purple\",\"5/54\"],[\"yellow\",\"5/54\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"gray": 0.01, "yellow": 0.01, "purple": 0.13, "black": 0.03, "blue": 0.01, "cyan": 0.17, "green": 0.63, "pink": 0.01}, "focus": {"false": 0.84, "true": 0.16}, "group": {"0": 0.03, "1": 0.67, "2": 0.3}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:185", "state_id": "known_chance_v2:train:185", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: brown=1, green=3, cyan=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "cyan": "The sampled ball is cyan", "brown": "The sampled ball is brown"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a green ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: brown", "The drawn color is one of: green", "The drawn color is one of: cyan"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.375, "cyan": 0.5, "brown": 0.125}, "focus": {"false": 0.625, "true": 0.375}, "group": {"0": 0.125, "1": 0.375, "2": 0.5}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"1/8\"],[\"cyan\",\"1/2\"],[\"green\",\"3/8\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0.81, "brown": 0.14, "green": 0.05}, "focus": {"false": 0.6699999999999999, "true": 0.33}, "group": {"0": 0.02, "1": 0.04, "2": 0.94}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:184", "state_id": "known_chance_v2:train:184", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: yellow=7, gray=2, black=6, white=9, pink=4, blue=6, cyan=5, orange=6, purple=4, brown=8, green=1, red=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"yellow": "The sampled ball is yellow", "gray": "The sampled ball is gray", "orange": "The sampled ball is orange", "purple": "The sampled ball is purple", "brown": "The sampled ball is brown", "blue": "The sampled ball is blue", "green": "The sampled ball is green", "red": "The sampled ball is red", "cyan": "The sampled ball is cyan", "pink": "The sampled ball is pink", "white": "The sampled ball is white", "black": "The sampled ball is black"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: yellow, white, cyan, brown", "The drawn color is one of: gray, pink, orange, green", "The drawn color is one of: black, blue, purple, red"]}}, "gold": {}, "gold_probs": {"outcome": {"yellow": 0.11475409836065574, "gray": 0.03278688524590164, "orange": 0.09836065573770492, "purple": 0.06557377049180328, "brown": 0.13114754098360656, "blue": 0.09836065573770492, "green": 0.01639344262295082, "red": 0.04918032786885246, "cyan": 0.08196721311475409, "pink": 0.06557377049180328, "white": 0.14754098360655737, "black": 0.09836065573770492}, "focus": {"false": 0.8852459016393442, "true": 0.11475409836065574}, "group": {"0": 0.47540983606557374, "1": 0.21311475409836067, "2": 0.3114754098360656}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"6/61\"],[\"blue\",\"6/61\"],[\"brown\",\"8/61\"],[\"cyan\",\"5/61\"],[\"gray\",\"2/61\"],[\"green\",\"1/61\"],[\"orange\",\"6/61\"],[\"pink\",\"4/61\"],[\"purple\",\"4/61\"],[\"red\",\"3/61\"],[\"white\",\"9/61\"],[\"yellow\",\"7/61\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0, "pink": 0, "green": 0.06, "orange": 0, "blue": 0, "red": 0, "cyan": 0, "white": 0.9, "yellow": 0.02, "brown": 0.01, "black": 0.01, "gray": 0}, "focus": {"false": 0.83, "true": 0.17}, "group": {"0": 0.99, "1": 0, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:186", "state_id": "known_chance_v2:train:186", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: gray=9, pink=6, black=4, white=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "gray": "The sampled ball is gray", "pink": "The sampled ball is pink", "white": "The sampled ball is white"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: gray, white", "The drawn color is one of: pink", "The drawn color is one of: black"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.16666666666666666, "gray": 0.375, "pink": 0.25, "white": 0.20833333333333334}, "focus": {"false": 0.8333333333333334, "true": 0.16666666666666666}, "group": {"0": 0.5833333333333334, "1": 0.25, "2": 0.16666666666666666}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/6\"],[\"gray\",\"3/8\"],[\"pink\",\"1/4\"],[\"white\",\"5/24\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0.01, "black": 0, "pink": 0, "gray": 0.99}, "focus": {"false": 0.83, "true": 0.17}, "group": {"0": 0.98, "1": 0.01, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:187", "state_id": "known_chance_v2:train:187", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: white=7, cyan=8, purple=7, brown=9, red=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"red": "The sampled ball is red", "purple": "The sampled ball is purple", "cyan": "The sampled ball is cyan", "brown": "The sampled ball is brown", "white": "The sampled ball is white"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white, brown", "The drawn color is one of: cyan, red", "The drawn color is one of: purple"]}}, "gold": {}, "gold_probs": {"outcome": {"red": 0.16216216216216217, "purple": 0.1891891891891892, "cyan": 0.21621621621621623, "brown": 0.24324324324324326, "white": 0.1891891891891892}, "focus": {"false": 0.7837837837837838, "true": 0.21621621621621623}, "group": {"0": 0.43243243243243246, "1": 0.3783783783783784, "2": 0.1891891891891892}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"9/37\"],[\"cyan\",\"8/37\"],[\"purple\",\"7/37\"],[\"red\",\"6/37\"],[\"white\",\"7/37\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0.15, "purple": 0.01, "red": 0.01, "brown": 0.78, "white": 0.05}, "focus": {"false": 0.77, "true": 0.23}, "group": {"0": 0.83, "1": 0.15, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:179", "state_id": "known_chance_v2:train:179", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: black=3, purple=1, pink=4, red=3, brown=5, orange=3, yellow=6, gray=4, white=9, blue=6, cyan=8, green=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "white": "The sampled ball is white", "green": "The sampled ball is green", "blue": "The sampled ball is blue", "gray": "The sampled ball is gray", "orange": "The sampled ball is orange", "brown": "The sampled ball is brown", "purple": "The sampled ball is purple", "pink": "The sampled ball is pink", "black": "The sampled ball is black", "red": "The sampled ball is red", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a purple ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: black, red, yellow, blue", "The drawn color is one of: purple, brown, gray, cyan", "The drawn color is one of: pink, orange, white, green"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.14035087719298245, "white": 0.15789473684210525, "green": 0.08771929824561403, "blue": 0.10526315789473684, "gray": 0.07017543859649122, "orange": 0.05263157894736842, "brown": 0.08771929824561403, "purple": 0.017543859649122806, "pink": 0.07017543859649122, "black": 0.05263157894736842, "red": 0.05263157894736842, "yellow": 0.10526315789473684}, "focus": {"false": 0.9824561403508771, "true": 0.017543859649122806}, "group": {"0": 0.3157894736842105, "1": 0.3157894736842105, "2": 0.3684210526315789}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/19\"],[\"blue\",\"2/19\"],[\"brown\",\"5/57\"],[\"cyan\",\"8/57\"],[\"gray\",\"4/57\"],[\"green\",\"5/57\"],[\"orange\",\"1/19\"],[\"pink\",\"4/57\"],[\"purple\",\"1/57\"],[\"red\",\"1/19\"],[\"white\",\"3/19\"],[\"yellow\",\"2/19\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"red": 0, "pink": 0, "yellow": 0.01, "white": 0.9, "blue": 0, "gray": 0, "purple": 0.03, "orange": 0, "cyan": 0.01, "black": 0.05, "green": 0, "brown": 0}, "focus": {"false": 0.94, "true": 0.06}, "group": {"0": 0.08, "1": 0.04, "2": 0.88}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:188", "state_id": "known_chance_v2:train:188", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: white=7, brown=9, pink=1, blue=5, gray=3, yellow=7, purple=4, red=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"purple": "The sampled ball is purple", "brown": "The sampled ball is brown", "blue": "The sampled ball is blue", "yellow": "The sampled ball is yellow", "pink": "The sampled ball is pink", "gray": "The sampled ball is gray", "red": "The sampled ball is red", "white": "The sampled ball is white"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a gray ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white, blue, purple", "The drawn color is one of: brown, gray, red", "The drawn color is one of: pink, yellow"]}}, "gold": {}, "gold_probs": {"outcome": {"purple": 0.10256410256410256, "brown": 0.23076923076923078, "blue": 0.1282051282051282, "yellow": 0.1794871794871795, "pink": 0.02564102564102564, "gray": 0.07692307692307693, "red": 0.07692307692307693, "white": 0.1794871794871795}, "focus": {"false": 0.9230769230769231, "true": 0.07692307692307693}, "group": {"0": 0.41025641025641024, "1": 0.38461538461538464, "2": 0.20512820512820512}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"5/39\"],[\"brown\",\"3/13\"],[\"gray\",\"1/13\"],[\"pink\",\"1/39\"],[\"purple\",\"4/39\"],[\"red\",\"1/13\"],[\"white\",\"7/39\"],[\"yellow\",\"7/39\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0, "brown": 0.92, "pink": 0.03, "gray": 0, "yellow": 0.01, "blue": 0, "white": 0.04, "red": 0}, "focus": {"false": 0.88, "true": 0.12}, "group": {"0": 0.19, "1": 0.79, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:189", "state_id": "known_chance_v2:train:189", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: yellow=7, orange=5, red=1, brown=6, gray=8, purple=2, white=6, green=2, black=6, blue=5, pink=3, cyan=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "white": "The sampled ball is white", "orange": "The sampled ball is orange", "yellow": "The sampled ball is yellow", "purple": "The sampled ball is purple", "black": "The sampled ball is black", "gray": "The sampled ball is gray", "pink": "The sampled ball is pink", "red": "The sampled ball is red", "cyan": "The sampled ball is cyan", "blue": "The sampled ball is blue", "brown": "The sampled ball is brown"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: yellow, brown, white, blue", "The drawn color is one of: orange, gray, green, pink", "The drawn color is one of: red, purple, black, cyan"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.03773584905660377, "white": 0.11320754716981132, "orange": 0.09433962264150944, "yellow": 0.1320754716981132, "purple": 0.03773584905660377, "black": 0.11320754716981132, "gray": 0.1509433962264151, "pink": 0.05660377358490566, "red": 0.018867924528301886, "cyan": 0.03773584905660377, "blue": 0.09433962264150944, "brown": 0.11320754716981132}, "focus": {"false": 0.8867924528301887, "true": 0.11320754716981132}, "group": {"0": 0.4528301886792453, "1": 0.33962264150943394, "2": 0.20754716981132076}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"6/53\"],[\"blue\",\"5/53\"],[\"brown\",\"6/53\"],[\"cyan\",\"2/53\"],[\"gray\",\"8/53\"],[\"green\",\"2/53\"],[\"orange\",\"5/53\"],[\"pink\",\"3/53\"],[\"purple\",\"2/53\"],[\"red\",\"1/53\"],[\"white\",\"6/53\"],[\"yellow\",\"7/53\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0, "pink": 0, "yellow": 0.06, "red": 0.03, "cyan": 0, "green": 0, "gray": 0.89, "brown": 0.01, "orange": 0, "purple": 0, "blue": 0.01, "black": 0}, "focus": {"false": 0.85, "true": 0.15}, "group": {"0": 0.87, "1": 0.1, "2": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:190", "state_id": "known_chance_v2:train:190", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: purple=8, yellow=6, white=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"yellow": "The sampled ball is yellow", "white": "The sampled ball is white", "purple": "The sampled ball is purple"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: purple", "The drawn color is one of: yellow", "The drawn color is one of: white"]}}, "gold": {}, "gold_probs": {"outcome": {"yellow": 0.3157894736842105, "white": 0.2631578947368421, "purple": 0.42105263157894735}, "focus": {"false": 0.6842105263157895, "true": 0.3157894736842105}, "group": {"0": 0.42105263157894735, "1": 0.3157894736842105, "2": 0.2631578947368421}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"purple\",\"8/19\"],[\"white\",\"5/19\"],[\"yellow\",\"6/19\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"yellow": 0.01, "purple": 0.99, "white": 0}, "focus": {"false": 0.73, "true": 0.27}, "group": {"0": 0.97, "1": 0.01, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:191", "state_id": "known_chance_v2:train:191", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: white=6, gray=4, blue=8, purple=9. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"purple": "The sampled ball is purple", "blue": "The sampled ball is blue", "white": "The sampled ball is white", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white, purple", "The drawn color is one of: gray", "The drawn color is one of: blue"]}}, "gold": {}, "gold_probs": {"outcome": {"purple": 0.3333333333333333, "blue": 0.2962962962962963, "white": 0.2222222222222222, "gray": 0.14814814814814814}, "focus": {"false": 0.7037037037037037, "true": 0.2962962962962963}, "group": {"0": 0.5555555555555556, "1": 0.14814814814814814, "2": 0.2962962962962963}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"8/27\"],[\"gray\",\"4/27\"],[\"purple\",\"1/3\"],[\"white\",\"2/9\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.71, "blue": 0.11, "white": 0.17, "gray": 0.01}, "focus": {"false": 0.77, "true": 0.23}, "group": {"0": 0.82, "1": 0.01, "2": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:192", "state_id": "known_chance_v2:train:192", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: red=4, orange=6, white=3, yellow=2, pink=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"pink": "The sampled ball is pink", "red": "The sampled ball is red", "yellow": "The sampled ball is yellow", "orange": "The sampled ball is orange", "white": "The sampled ball is white"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: red, yellow", "The drawn color is one of: orange, pink", "The drawn color is one of: white"]}}, "gold": {}, "gold_probs": {"outcome": {"pink": 0.25, "red": 0.2, "yellow": 0.1, "orange": 0.3, "white": 0.15}, "focus": {"false": 0.9, "true": 0.1}, "group": {"0": 0.30000000000000004, "1": 0.55, "2": 0.15}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"orange\",\"3/10\"],[\"pink\",\"1/4\"],[\"red\",\"1/5\"],[\"white\",\"3/20\"],[\"yellow\",\"1/10\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.01, "orange": 0.92, "yellow": 0.02, "red": 0.05, "white": 0}, "focus": {"false": 0.87, "true": 0.13}, "group": {"0": 0.01, "1": 0.98, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:193", "state_id": "known_chance_v2:train:193", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: white=4, red=2, orange=3, yellow=7, cyan=3, pink=1, brown=3, blue=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"red": "The sampled ball is red", "blue": "The sampled ball is blue", "white": "The sampled ball is white", "pink": "The sampled ball is pink", "yellow": "The sampled ball is yellow", "orange": "The sampled ball is orange", "brown": "The sampled ball is brown", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a red ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white, yellow, brown", "The drawn color is one of: red, cyan, blue", "The drawn color is one of: orange, pink"]}}, "gold": {}, "gold_probs": {"outcome": {"red": 0.08, "blue": 0.08, "white": 0.16, "pink": 0.04, "yellow": 0.28, "orange": 0.12, "brown": 0.12, "cyan": 0.12}, "focus": {"false": 0.92, "true": 0.08}, "group": {"0": 0.56, "1": 0.28, "2": 0.16}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"2/25\"],[\"brown\",\"3/25\"],[\"cyan\",\"3/25\"],[\"orange\",\"3/25\"],[\"pink\",\"1/25\"],[\"red\",\"2/25\"],[\"white\",\"4/25\"],[\"yellow\",\"7/25\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0, "white": 0.06, "pink": 0.01, "yellow": 0.93, "blue": 0, "brown": 0, "red": 0, "orange": 0}, "focus": {"false": 0.85, "true": 0.15}, "group": {"0": 0.99, "1": 0.01, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:194", "state_id": "known_chance_v2:train:194", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: brown=2, black=6, white=6, pink=7, orange=4, blue=7, purple=1, red=8, cyan=2, yellow=7, gray=3, green=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "gray": "The sampled ball is gray", "orange": "The sampled ball is orange", "purple": "The sampled ball is purple", "cyan": "The sampled ball is cyan", "red": "The sampled ball is red", "blue": "The sampled ball is blue", "pink": "The sampled ball is pink", "white": "The sampled ball is white", "black": "The sampled ball is black", "brown": "The sampled ball is brown", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: brown, pink, purple, yellow", "The drawn color is one of: black, orange, red, gray", "The drawn color is one of: white, blue, cyan, green"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.03636363636363636, "gray": 0.05454545454545454, "orange": 0.07272727272727272, "purple": 0.01818181818181818, "cyan": 0.03636363636363636, "red": 0.14545454545454545, "blue": 0.12727272727272726, "pink": 0.12727272727272726, "white": 0.10909090909090909, "black": 0.10909090909090909, "brown": 0.03636363636363636, "yellow": 0.12727272727272726}, "focus": {"false": 0.8909090909090909, "true": 0.10909090909090909}, "group": {"0": 0.3090909090909091, "1": 0.3818181818181818, "2": 0.3090909090909091}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"6/55\"],[\"blue\",\"7/55\"],[\"brown\",\"2/55\"],[\"cyan\",\"2/55\"],[\"gray\",\"3/55\"],[\"green\",\"2/55\"],[\"orange\",\"4/55\"],[\"pink\",\"7/55\"],[\"purple\",\"1/55\"],[\"red\",\"8/55\"],[\"white\",\"6/55\"],[\"yellow\",\"7/55\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"blue": 0.01, "red": 0.9, "white": 0, "orange": 0, "green": 0, "black": 0.02, "gray": 0, "yellow": 0.01, "cyan": 0, "purple": 0.02, "pink": 0.02, "brown": 0.02}, "focus": {"false": 0.83, "true": 0.17}, "group": {"0": 0.1, "1": 0.79, "2": 0.11}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:196", "state_id": "known_chance_v2:train:196", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: green=4, black=9, orange=2, white=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "green": "The sampled ball is green", "orange": "The sampled ball is orange", "white": "The sampled ball is white"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a white ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: green, white", "The drawn color is one of: black", "The drawn color is one of: orange"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.42857142857142855, "green": 0.19047619047619047, "orange": 0.09523809523809523, "white": 0.2857142857142857}, "focus": {"false": 0.7142857142857143, "true": 0.2857142857142857}, "group": {"0": 0.47619047619047616, "1": 0.42857142857142855, "2": 0.09523809523809523}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"3/7\"],[\"green\",\"4/21\"],[\"orange\",\"2/21\"],[\"white\",\"2/7\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0.01, "orange": 0, "green": 0.02, "black": 0.97}, "focus": {"false": 0.79, "true": 0.21}, "group": {"0": 0.03, "1": 0.97, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:195", "state_id": "known_chance_v2:train:195", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: purple=1, cyan=5, red=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"purple": "The sampled ball is purple", "cyan": "The sampled ball is cyan", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a red ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: purple", "The drawn color is one of: cyan", "The drawn color is one of: red"]}}, "gold": {}, "gold_probs": {"outcome": {"purple": 0.09090909090909091, "cyan": 0.45454545454545453, "red": 0.45454545454545453}, "focus": {"false": 0.5454545454545454, "true": 0.45454545454545453}, "group": {"0": 0.09090909090909091, "1": 0.45454545454545453, "2": 0.45454545454545453}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"5/11\"],[\"purple\",\"1/11\"],[\"red\",\"5/11\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"red": 0.04, "cyan": 0.73, "purple": 0.23}, "focus": {"false": 0.69, "true": 0.31}, "group": {"0": 0.03, "1": 0.86, "2": 0.11}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:198", "state_id": "known_chance_v2:train:198", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: gray=8, blue=1, brown=8, yellow=1, cyan=2, orange=5, red=7, pink=9. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"brown": "The sampled ball is brown", "blue": "The sampled ball is blue", "pink": "The sampled ball is pink", "cyan": "The sampled ball is cyan", "gray": "The sampled ball is gray", "red": "The sampled ball is red", "orange": "The sampled ball is orange", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: gray, yellow, red", "The drawn color is one of: blue, cyan, pink", "The drawn color is one of: brown, orange"]}}, "gold": {}, "gold_probs": {"outcome": {"brown": 0.1951219512195122, "blue": 0.024390243902439025, "pink": 0.21951219512195122, "cyan": 0.04878048780487805, "gray": 0.1951219512195122, "red": 0.17073170731707318, "orange": 0.12195121951219512, "yellow": 0.024390243902439025}, "focus": {"false": 0.975609756097561, "true": 0.024390243902439025}, "group": {"0": 0.3902439024390244, "1": 0.2926829268292683, "2": 0.3170731707317073}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"1/41\"],[\"brown\",\"8/41\"],[\"cyan\",\"2/41\"],[\"gray\",\"8/41\"],[\"orange\",\"5/41\"],[\"pink\",\"9/41\"],[\"red\",\"7/41\"],[\"yellow\",\"1/41\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0.01, "gray": 0.05, "brown": 0.01, "orange": 0.01, "yellow": 0.01, "red": 0.01, "pink": 0.89, "blue": 0.01}, "focus": {"false": 0.94, "true": 0.06}, "group": {"0": 0.27, "1": 0.29, "2": 0.44}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:197", "state_id": "known_chance_v2:train:197", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: pink=8, black=1, red=4, cyan=8, purple=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "black": "The sampled ball is black", "purple": "The sampled ball is purple", "pink": "The sampled ball is pink", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a purple ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: pink, cyan", "The drawn color is one of: black, purple", "The drawn color is one of: red"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.34782608695652173, "black": 0.043478260869565216, "purple": 0.08695652173913043, "pink": 0.34782608695652173, "red": 0.17391304347826086}, "focus": {"false": 0.9130434782608696, "true": 0.08695652173913043}, "group": {"0": 0.6956521739130435, "1": 0.13043478260869565, "2": 0.17391304347826086}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/23\"],[\"cyan\",\"8/23\"],[\"pink\",\"8/23\"],[\"purple\",\"2/23\"],[\"red\",\"4/23\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"red": 0, "cyan": 0.12, "purple": 0, "black": 0.01, "pink": 0.87}, "focus": {"false": 0.91, "true": 0.09}, "group": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:train:199", "state_id": "known_chance_v2:train:199", "family_id": "known_chance_v2", "split": "train", "state": "Random-draw machine: exactly one ball is drawn uniformly from a bag. Counts: gray=8, pink=8, purple=3, black=5, blue=4, white=7, cyan=3, orange=2, yellow=8, red=3, brown=3, green=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"red": "The sampled ball is red", "purple": "The sampled ball is purple", "green": "The sampled ball is green", "blue": "The sampled ball is blue", "black": "The sampled ball is black", "brown": "The sampled ball is brown", "cyan": "The sampled ball is cyan", "yellow": "The sampled ball is yellow", "gray": "The sampled ball is gray", "white": "The sampled ball is white", "pink": "The sampled ball is pink", "orange": "The sampled ball is orange"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: gray, black, cyan, red", "The drawn color is one of: pink, blue, orange, brown", "The drawn color is one of: purple, white, yellow, green"]}}, "gold": {}, "gold_probs": {"outcome": {"red": 0.04838709677419355, "purple": 0.04838709677419355, "green": 0.12903225806451613, "blue": 0.06451612903225806, "black": 0.08064516129032258, "brown": 0.04838709677419355, "cyan": 0.04838709677419355, "yellow": 0.12903225806451613, "gray": 0.12903225806451613, "white": 0.11290322580645161, "pink": 0.12903225806451613, "orange": 0.03225806451612903}, "focus": {"false": 0.8709677419354839, "true": 0.12903225806451613}, "group": {"0": 0.3064516129032258, "1": 0.27419354838709675, "2": 0.4193548387096774}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"5/62\"],[\"blue\",\"2/31\"],[\"brown\",\"3/62\"],[\"cyan\",\"3/62\"],[\"gray\",\"4/31\"],[\"green\",\"4/31\"],[\"orange\",\"1/31\"],[\"pink\",\"4/31\"],[\"purple\",\"3/62\"],[\"red\",\"3/62\"],[\"white\",\"7/62\"],[\"yellow\",\"4/31\"]]", "template_id": "known_chance_v2:train"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"brown": 0, "cyan": 0, "yellow": 0.03, "white": 0.02, "gray": 0.67, "purple": 0.01, "black": 0.01, "green": 0.06, "blue": 0.01, "red": 0.01, "pink": 0.03, "orange": 0.15}, "focus": {"false": 0.84, "true": 0.16}, "group": {"0": 0.23, "1": 0.07, "2": 0.7}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:dev:1", "state_id": "smart_home_v2:dev:1", "family_id": "smart_home_v2", "split": "dev", "state": "Device command received: room garden, device speaker, action on. Permission no. Occupancy 0. Hour 2.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom lamp", "unit_1": "garden speaker", "unit_2": "balcony heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garden\",\"speaker\",false,0]", "template_id": "smart_home_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_2": 0, "unit_1": 1}, "execute": {"false": 0.43999999999999995, "true": 0.56}, "risk": {"0": 0.49, "1": 0.1, "2": 0.41}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:dev:0", "state_id": "smart_home_v2:dev:0", "family_id": "smart_home_v2", "split": "dev", "state": "Device command received: room attic, device fan, action on. Permission yes. Occupancy 0. Hour 13.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library fan", "unit_1": "attic fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"attic\",\"fan\",true,0]", "template_id": "smart_home_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 1, "unit_0": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 0.99, "1": 0.01, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:dev:2", "state_id": "smart_home_v2:dev:2", "family_id": "smart_home_v2", "split": "dev", "state": "Device command received: room library, device speaker, action on. Permission yes. Occupancy 3. Hour 13.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library speaker", "unit_1": "garden speaker", "unit_2": "garage fan", "unit_3": "attic fan", "unit_4": "office lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"library\",\"speaker\",true,3]", "template_id": "smart_home_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_2": 0, "unit_1": 0, "unit_4": 0, "unit_3": 0}, "execute": {"false": 0.040000000000000036, "true": 0.96}, "risk": {"0": 0.99, "1": 0, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:dev:7", "state_id": "smart_home_v2:dev:7", "family_id": "smart_home_v2", "split": "dev", "state": "Device command received: room bedroom, device speaker, action off. Permission no. Occupancy 2. Hour 20.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony heater", "unit_1": "bedroom speaker", "unit_2": "garden fan", "unit_3": "bathroom heater", "unit_4": "living room heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0, "unit_3": 0, "unit_4": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bedroom\",\"speaker\",false,2]", "template_id": "smart_home_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_2": 0, "unit_4": 0, "unit_3": 0, "unit_1": 1}, "execute": {"false": 0.6799999999999999, "true": 0.32}, "risk": {"0": 0.64, "1": 0, "2": 0.36}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:dev:3", "state_id": "smart_home_v2:dev:3", "family_id": "smart_home_v2", "split": "dev", "state": "Device command received: room office, device lamp, action on. Permission no. Occupancy 3. Hour 4.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library fan", "unit_1": "kitchen fan", "unit_2": "hall heater", "unit_3": "kitchen heater", "unit_4": "garden lamp", "unit_5": "office lamp", "unit_6": "living room fan", "unit_7": "kitchen lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_5", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 1, "unit_6": 0, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"office\",\"lamp\",false,3]", "template_id": "smart_home_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_6": 0, "unit_5": 1, "unit_2": 0, "unit_7": 0, "unit_3": 0, "unit_4": 0}, "execute": {"false": 0.39, "true": 0.61}, "risk": {"0": 0.85, "1": 0, "2": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:dev:6", "state_id": "smart_home_v2:dev:6", "family_id": "smart_home_v2", "split": "dev", "state": "Device command received: room dining room, device heater, action off. Permission yes. Occupancy 3. Hour 5.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "hall speaker", "unit_1": "dining room heater", "unit_2": "kitchen lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"dining room\",\"heater\",true,3]", "template_id": "smart_home_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 1, "unit_2": 0, "unit_0": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:dev:5", "state_id": "smart_home_v2:dev:5", "family_id": "smart_home_v2", "split": "dev", "state": "Device command received: room garden, device speaker, action off. Permission yes. Occupancy 1. Hour 18.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garden speaker", "unit_1": "garden lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garden\",\"speaker\",true,1]", "template_id": "smart_home_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_0": 1}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 0.99, "1": 0, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:dev:9", "state_id": "smart_home_v2:dev:9", "family_id": "smart_home_v2", "split": "dev", "state": "Device command received: room garden, device heater, action on. Permission yes. Occupancy 3. Hour 20.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "dining room lamp", "unit_1": "garden speaker", "unit_2": "balcony speaker", "unit_3": "bedroom lamp", "unit_4": "dining room heater", "unit_5": "garage heater", "unit_6": "garden heater", "unit_7": "living room lamp", "unit_8": "bathroom lamp", "unit_9": "library speaker", "unit_10": "balcony fan", "unit_11": "office lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_6", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 1, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garden\",\"heater\",true,3]", "template_id": "smart_home_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_2": 0, "unit_4": 0, "unit_7": 0, "unit_5": 0, "unit_10": 0, "unit_1": 0, "unit_11": 0, "unit_8": 0, "unit_6": 1, "unit_3": 0, "unit_9": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:dev:8", "state_id": "smart_home_v2:dev:8", "family_id": "smart_home_v2", "split": "dev", "state": "Device command received: room kitchen, device fan, action on. Permission no. Occupancy 3. Hour 15.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "kitchen lamp", "unit_1": "garden speaker", "unit_2": "office lamp", "unit_3": "bathroom lamp", "unit_4": "office fan", "unit_5": "dining room speaker", "unit_6": "attic lamp", "unit_7": "kitchen fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_7", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"kitchen\",\"fan\",false,3]", "template_id": "smart_home_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_4": 0, "unit_3": 0, "unit_7": 1, "unit_1": 0, "unit_6": 0, "unit_0": 0, "unit_5": 0}, "execute": {"false": 0.5900000000000001, "true": 0.41}, "risk": {"0": 0.79, "1": 0, "2": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:dev:11", "state_id": "smart_home_v2:dev:11", "family_id": "smart_home_v2", "split": "dev", "state": "Device command received: room balcony, device lamp, action off. Permission yes. Occupancy 3. Hour 11.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony lamp", "unit_1": "office heater", "unit_2": "garage speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"balcony\",\"lamp\",true,3]", "template_id": "smart_home_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_0": 1, "unit_2": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:dev:4", "state_id": "smart_home_v2:dev:4", "family_id": "smart_home_v2", "split": "dev", "state": "Device command received: room garden, device speaker, action off. Permission no. Occupancy 1. Hour 20.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room speaker", "unit_1": "garage fan", "unit_2": "dining room fan", "unit_3": "dining room speaker", "unit_4": "garage lamp", "unit_5": "office heater", "unit_6": "balcony lamp", "unit_7": "office speaker", "unit_8": "garden speaker", "unit_9": "garden lamp", "unit_10": "kitchen lamp", "unit_11": "hall heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_8", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 1, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garden\",\"speaker\",false,1]", "template_id": "smart_home_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_2": 0, "unit_8": 1, "unit_5": 0, "unit_9": 0, "unit_0": 0, "unit_11": 0, "unit_7": 0, "unit_4": 0, "unit_10": 0, "unit_6": 0, "unit_3": 0}, "execute": {"false": 0.8, "true": 0.2}, "risk": {"0": 0.51, "1": 0, "2": 0.49}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:dev:10", "state_id": "smart_home_v2:dev:10", "family_id": "smart_home_v2", "split": "dev", "state": "Device command received: room garden, device fan, action on. Permission no. Occupancy 1. Hour 14.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garden fan", "unit_1": "library heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garden\",\"fan\",false,1]", "template_id": "smart_home_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0.25, "true": 0.75}, "risk": {"0": 0.96, "1": 0, "2": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:dev:13", "state_id": "smart_home_v2:dev:13", "family_id": "smart_home_v2", "split": "dev", "state": "Device command received: room office, device lamp, action on. Permission no. Occupancy 1. Hour 13.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "office lamp", "unit_1": "living room speaker", "unit_2": "garden heater", "unit_3": "office speaker", "unit_4": "library lamp", "unit_5": "library heater", "unit_6": "balcony fan", "unit_7": "hall heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"office\",\"lamp\",false,1]", "template_id": "smart_home_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_1": 0, "unit_6": 0, "unit_7": 0, "unit_0": 1, "unit_3": 0, "unit_5": 0, "unit_4": 0}, "execute": {"false": 0.19999999999999996, "true": 0.8}, "risk": {"0": 0.93, "1": 0, "2": 0.07}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:dev:14", "state_id": "smart_home_v2:dev:14", "family_id": "smart_home_v2", "split": "dev", "state": "Device command received: room bathroom, device fan, action off. Permission yes. Occupancy 2. Hour 2.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "office heater", "unit_1": "kitchen lamp", "unit_2": "garage fan", "unit_3": "office lamp", "unit_4": "bathroom fan", "unit_5": "kitchen speaker", "unit_6": "kitchen fan", "unit_7": "living room lamp", "unit_8": "attic lamp", "unit_9": "hall speaker", "unit_10": "dining room speaker", "unit_11": "bathroom speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_4", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 1, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bathroom\",\"fan\",true,2]", "template_id": "smart_home_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_6": 0, "unit_2": 0, "unit_1": 0, "unit_7": 0, "unit_8": 0, "unit_0": 0, "unit_5": 0, "unit_3": 0, "unit_11": 0, "unit_4": 1, "unit_9": 0, "unit_10": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:dev:12", "state_id": "smart_home_v2:dev:12", "family_id": "smart_home_v2", "split": "dev", "state": "Device command received: room kitchen, device heater, action off. Permission no. Occupancy 2. Hour 16.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "hall fan", "unit_1": "garage speaker", "unit_2": "bathroom fan", "unit_3": "living room lamp", "unit_4": "kitchen heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_4", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"kitchen\",\"heater\",false,2]", "template_id": "smart_home_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_1": 0, "unit_4": 1, "unit_0": 0, "unit_3": 0}, "execute": {"false": 0.79, "true": 0.21}, "risk": {"0": 0.71, "1": 0, "2": 0.29}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:dev:16", "state_id": "smart_home_v2:dev:16", "family_id": "smart_home_v2", "split": "dev", "state": "Device command received: room garden, device fan, action on. Permission yes. Occupancy 2. Hour 16.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garden fan", "unit_1": "attic speaker", "unit_2": "office heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garden\",\"fan\",true,2]", "template_id": "smart_home_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_1": 0, "unit_0": 1}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:dev:15", "state_id": "smart_home_v2:dev:15", "family_id": "smart_home_v2", "split": "dev", "state": "Device command received: room bathroom, device heater, action on. Permission yes. Occupancy 1. Hour 13.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bathroom heater", "unit_1": "living room heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bathroom\",\"heater\",true,1]", "template_id": "smart_home_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:dev:19", "state_id": "smart_home_v2:dev:19", "family_id": "smart_home_v2", "split": "dev", "state": "Device command received: room bedroom, device lamp, action on. Permission yes. Occupancy 0. Hour 18.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room heater", "unit_1": "library speaker", "unit_2": "garage heater", "unit_3": "bedroom lamp", "unit_4": "dining room speaker", "unit_5": "living room fan", "unit_6": "bathroom heater", "unit_7": "office lamp", "unit_8": "library heater", "unit_9": "hall heater", "unit_10": "balcony heater", "unit_11": "garden speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_3", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 1, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bedroom\",\"lamp\",true,0]", "template_id": "smart_home_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_0": 0, "unit_11": 0, "unit_4": 0, "unit_3": 1, "unit_9": 0, "unit_7": 0, "unit_1": 0, "unit_6": 0, "unit_10": 0, "unit_8": 0, "unit_5": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 0.99, "1": 0, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:dev:17", "state_id": "smart_home_v2:dev:17", "family_id": "smart_home_v2", "split": "dev", "state": "Device command received: room attic, device heater, action off. Permission yes. Occupancy 2. Hour 1.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "attic heater", "unit_1": "living room fan", "unit_2": "office heater", "unit_3": "living room speaker", "unit_4": "garage heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"attic\",\"heater\",true,2]", "template_id": "smart_home_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_2": 0, "unit_4": 0, "unit_0": 1, "unit_3": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:dev:18", "state_id": "smart_home_v2:dev:18", "family_id": "smart_home_v2", "split": "dev", "state": "Device command received: room dining room, device speaker, action off. Permission no. Occupancy 3. Hour 10.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "kitchen speaker", "unit_1": "dining room speaker", "unit_2": "bathroom fan", "unit_3": "garage heater", "unit_4": "balcony heater", "unit_5": "living room lamp", "unit_6": "bedroom fan", "unit_7": "hall speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"dining room\",\"speaker\",false,3]", "template_id": "smart_home_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_6": 0, "unit_7": 0, "unit_5": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0}, "execute": {"false": 0.79, "true": 0.21}, "risk": {"0": 0.6, "1": 0, "2": 0.4}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:dev:20", "state_id": "smart_home_v2:dev:20", "family_id": "smart_home_v2", "split": "dev", "state": "Device command received: room balcony, device heater, action on. Permission no. Occupancy 2. Hour 23.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony heater", "unit_1": "attic lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"balcony\",\"heater\",false,2]", "template_id": "smart_home_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_0": 1}, "execute": {"false": 0.18000000000000005, "true": 0.82}, "risk": {"0": 0.92, "1": 0, "2": 0.08}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:dev:21", "state_id": "smart_home_v2:dev:21", "family_id": "smart_home_v2", "split": "dev", "state": "Device command received: room living room, device fan, action on. Permission no. Occupancy 3. Hour 23.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room fan", "unit_1": "balcony heater", "unit_2": "garage speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"living room\",\"fan\",false,3]", "template_id": "smart_home_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_2": 0, "unit_1": 0}, "execute": {"false": 0.5700000000000001, "true": 0.43}, "risk": {"0": 0.76, "1": 0, "2": 0.24}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:dev:22", "state_id": "smart_home_v2:dev:22", "family_id": "smart_home_v2", "split": "dev", "state": "Device command received: room office, device speaker, action on. Permission no. Occupancy 3. Hour 16.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library speaker", "unit_1": "balcony lamp", "unit_2": "office speaker", "unit_3": "hall speaker", "unit_4": "bathroom heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1, "unit_3": 0, "unit_4": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"office\",\"speaker\",false,3]", "template_id": "smart_home_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_4": 0, "unit_1": 0, "unit_2": 1, "unit_0": 0, "unit_3": 0}, "execute": {"false": 0.36, "true": 0.64}, "risk": {"0": 0.86, "1": 0, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:dev:23", "state_id": "smart_home_v2:dev:23", "family_id": "smart_home_v2", "split": "dev", "state": "Device command received: room office, device fan, action off. Permission no. Occupancy 2. Hour 1.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garden lamp", "unit_1": "balcony speaker", "unit_2": "bedroom lamp", "unit_3": "dining room speaker", "unit_4": "attic heater", "unit_5": "bathroom lamp", "unit_6": "office fan", "unit_7": "kitchen speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_6", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 1, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"office\",\"fan\",false,2]", "template_id": "smart_home_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_4": 0, "unit_7": 0, "unit_2": 0, "unit_5": 0, "unit_1": 0, "unit_0": 0, "unit_3": 0, "unit_6": 1}, "execute": {"false": 0.76, "true": 0.24}, "risk": {"0": 0.56, "1": 0, "2": 0.44}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:dev:24", "state_id": "smart_home_v2:dev:24", "family_id": "smart_home_v2", "split": "dev", "state": "Device command received: room garage, device speaker, action on. Permission yes. Occupancy 1. Hour 0.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "dining room speaker", "unit_1": "library speaker", "unit_2": "attic speaker", "unit_3": "dining room heater", "unit_4": "balcony fan", "unit_5": "garden fan", "unit_6": "hall heater", "unit_7": "garage speaker", "unit_8": "bedroom speaker", "unit_9": "library fan", "unit_10": "bathroom heater", "unit_11": "hall lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_7", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 1, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garage\",\"speaker\",true,1]", "template_id": "smart_home_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_4": 0, "unit_10": 0, "unit_8": 0, "unit_7": 1, "unit_11": 0, "unit_3": 0, "unit_9": 0, "unit_1": 0, "unit_5": 0, "unit_6": 0, "unit_2": 0}, "execute": {"false": 0.040000000000000036, "true": 0.96}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:dev:25", "state_id": "smart_home_v2:dev:25", "family_id": "smart_home_v2", "split": "dev", "state": "Device command received: room office, device speaker, action on. Permission no. Occupancy 0. Hour 11.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "attic lamp", "unit_1": "office speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"office\",\"speaker\",false,0]", "template_id": "smart_home_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 1, "unit_0": 0}, "execute": {"false": 0.45999999999999996, "true": 0.54}, "risk": {"0": 0.56, "1": 0.06, "2": 0.38}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:dev:27", "state_id": "smart_home_v2:dev:27", "family_id": "smart_home_v2", "split": "dev", "state": "Device command received: room dining room, device heater, action on. Permission no. Occupancy 2. Hour 2.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "dining room fan", "unit_1": "attic heater", "unit_2": "hall lamp", "unit_3": "dining room heater", "unit_4": "kitchen fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_3", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 1, "unit_4": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"dining room\",\"heater\",false,2]", "template_id": "smart_home_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_0": 0, "unit_3": 1, "unit_4": 0, "unit_1": 0}, "execute": {"false": 0.13, "true": 0.87}, "risk": {"0": 0.96, "1": 0, "2": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:dev:28", "state_id": "smart_home_v2:dev:28", "family_id": "smart_home_v2", "split": "dev", "state": "Device command received: room dining room, device lamp, action off. Permission no. Occupancy 3. Hour 7.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garden speaker", "unit_1": "library fan", "unit_2": "dining room fan", "unit_3": "bathroom lamp", "unit_4": "dining room lamp", "unit_5": "garden fan", "unit_6": "hall speaker", "unit_7": "attic fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_4", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 1, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"dining room\",\"lamp\",false,3]", "template_id": "smart_home_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_4": 1, "unit_7": 0, "unit_3": 0, "unit_2": 0, "unit_1": 0, "unit_6": 0, "unit_5": 0}, "execute": {"false": 0.87, "true": 0.13}, "risk": {"0": 0.54, "1": 0, "2": 0.46}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:dev:29", "state_id": "smart_home_v2:dev:29", "family_id": "smart_home_v2", "split": "dev", "state": "Device command received: room kitchen, device fan, action off. Permission yes. Occupancy 2. Hour 8.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "hall lamp", "unit_1": "garage speaker", "unit_2": "kitchen fan", "unit_3": "attic heater", "unit_4": "attic lamp", "unit_5": "hall heater", "unit_6": "office heater", "unit_7": "dining room lamp", "unit_8": "attic speaker", "unit_9": "garage lamp", "unit_10": "garden fan", "unit_11": "garage fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"kitchen\",\"fan\",true,2]", "template_id": "smart_home_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_6": 0, "unit_8": 0, "unit_10": 0, "unit_3": 0, "unit_2": 1, "unit_1": 0, "unit_9": 0, "unit_11": 0, "unit_5": 0, "unit_0": 0, "unit_4": 0, "unit_7": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:dev:30", "state_id": "smart_home_v2:dev:30", "family_id": "smart_home_v2", "split": "dev", "state": "Device command received: room living room, device lamp, action on. Permission no. Occupancy 0. Hour 2.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom heater", "unit_1": "living room lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"living room\",\"lamp\",false,0]", "template_id": "smart_home_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 0.47, "true": 0.53}, "risk": {"0": 0.78, "1": 0.01, "2": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:dev:26", "state_id": "smart_home_v2:dev:26", "family_id": "smart_home_v2", "split": "dev", "state": "Device command received: room office, device fan, action on. Permission yes. Occupancy 1. Hour 0.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony heater", "unit_1": "balcony speaker", "unit_2": "office fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"office\",\"fan\",true,1]", "template_id": "smart_home_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:dev:0", "state_id": "catalog_lookup_v2:dev:0", "family_id": "catalog_lookup_v2", "split": "dev", "state": "Warehouse audit: item=SKU-00151; color=black; size=small; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "green, large", "candidate_1": "black, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00151\",\"black\",\"small\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:dev:2", "state_id": "catalog_lookup_v2:dev:2", "family_id": "catalog_lookup_v2", "split": "dev", "state": "Warehouse audit: item=SKU-00072; color=yellow; size=small; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "gray, medium", "candidate_1": "brown, small", "candidate_2": "yellow, small", "candidate_3": "white, large", "candidate_4": "purple, medium", "candidate_5": "purple, large", "candidate_6": "pink, large", "candidate_7": "pink, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_2", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 1, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00072\",\"yellow\",\"small\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_5": 0, "candidate_0": 0, "candidate_2": 1, "candidate_4": 0, "candidate_1": 0, "candidate_6": 0, "candidate_3": 0, "candidate_7": 0}, "ready": {"false": 0.72, "true": 0.28}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:dev:31", "state_id": "smart_home_v2:dev:31", "family_id": "smart_home_v2", "split": "dev", "state": "Device command received: room bedroom, device lamp, action off. Permission yes. Occupancy 1. Hour 17.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom fan", "unit_1": "library lamp", "unit_2": "bedroom lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bedroom\",\"lamp\",true,1]", "template_id": "smart_home_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 1, "unit_1": 0, "unit_0": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:dev:1", "state_id": "catalog_lookup_v2:dev:1", "family_id": "catalog_lookup_v2", "split": "dev", "state": "Warehouse audit: item=SKU-00009; color=red; size=medium; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "gray, large", "candidate_1": "red, medium", "candidate_2": "blue, large", "candidate_3": "white, large", "candidate_4": "black, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00009\",\"red\",\"medium\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 1, "candidate_4": 0, "candidate_3": 0, "candidate_0": 0, "candidate_2": 0}, "ready": {"false": 0.75, "true": 0.25}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:dev:3", "state_id": "catalog_lookup_v2:dev:3", "family_id": "catalog_lookup_v2", "split": "dev", "state": "Warehouse audit: item=SKU-00167; color=black; size=large; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "gray, large", "candidate_1": "yellow, small", "candidate_2": "green, large", "candidate_3": "green, small", "candidate_4": "blue, small", "candidate_5": "blue, medium", "candidate_6": "red, small", "candidate_7": "pink, medium", "candidate_8": "yellow, medium", "candidate_9": "cyan, large", "candidate_10": "black, large", "candidate_11": "gray, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_10", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 1, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00167\",\"black\",\"large\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_4": 0, "candidate_11": 0, "candidate_10": 1, "candidate_6": 0, "candidate_3": 0, "candidate_5": 0, "candidate_0": 0, "candidate_9": 0, "candidate_8": 0, "candidate_2": 0, "candidate_7": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:dev:6", "state_id": "catalog_lookup_v2:dev:6", "family_id": "catalog_lookup_v2", "split": "dev", "state": "Warehouse audit: item=SKU-00133; color=white; size=medium; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "gray, small", "candidate_1": "blue, medium", "candidate_2": "orange, medium", "candidate_3": "white, medium", "candidate_4": "blue, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_3", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 1, "candidate_4": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00133\",\"white\",\"medium\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_3": 1, "candidate_1": 0, "candidate_2": 0, "candidate_0": 0, "candidate_4": 0}, "ready": {"false": 0.12, "true": 0.88}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:dev:5", "state_id": "catalog_lookup_v2:dev:5", "family_id": "catalog_lookup_v2", "split": "dev", "state": "Warehouse audit: item=SKU-00105; color=purple; size=medium; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "cyan, medium", "candidate_1": "purple, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00105\",\"purple\",\"medium\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 0.74, "true": 0.26}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:dev:8", "state_id": "catalog_lookup_v2:dev:8", "family_id": "catalog_lookup_v2", "split": "dev", "state": "Warehouse audit: item=SKU-00176; color=orange; size=medium; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, small", "candidate_1": "purple, large", "candidate_2": "yellow, small", "candidate_3": "orange, medium", "candidate_4": "green, large", "candidate_5": "yellow, medium", "candidate_6": "brown, large", "candidate_7": "brown, medium", "candidate_8": "yellow, large", "candidate_9": "black, small", "candidate_10": "orange, large", "candidate_11": "gray, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_3", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 1, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00176\",\"orange\",\"medium\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_5": 0, "candidate_0": 0, "candidate_2": 0, "candidate_9": 0, "candidate_10": 0, "candidate_4": 0, "candidate_8": 0, "candidate_11": 0, "candidate_1": 0, "candidate_3": 1, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 0.71, "true": 0.29}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:dev:10", "state_id": "catalog_lookup_v2:dev:10", "family_id": "catalog_lookup_v2", "split": "dev", "state": "Warehouse audit: item=SKU-00227; color=brown; size=medium; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "cyan, large", "candidate_1": "brown, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00227\",\"brown\",\"medium\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 0.88, "true": 0.12}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:dev:11", "state_id": "catalog_lookup_v2:dev:11", "family_id": "catalog_lookup_v2", "split": "dev", "state": "Warehouse audit: item=SKU-00109; color=purple; size=medium; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "brown, small", "candidate_1": "pink, medium", "candidate_2": "purple, medium", "candidate_3": "yellow, small", "candidate_4": "red, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_2", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 1, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00109\",\"purple\",\"medium\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 0, "candidate_4": 0, "candidate_3": 0, "candidate_1": 0, "candidate_2": 1}, "ready": {"false": 0.09999999999999998, "true": 0.9}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:dev:12", "state_id": "catalog_lookup_v2:dev:12", "family_id": "catalog_lookup_v2", "split": "dev", "state": "Warehouse audit: item=SKU-00117; color=purple; size=large; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, large", "candidate_1": "pink, medium", "candidate_2": "black, small", "candidate_3": "yellow, large", "candidate_4": "green, large", "candidate_5": "orange, large", "candidate_6": "white, medium", "candidate_7": "red, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00117\",\"purple\",\"large\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_5": 0, "candidate_6": 0, "candidate_2": 0, "candidate_4": 0, "candidate_0": 1, "candidate_7": 0, "candidate_3": 0}, "ready": {"false": 0.10999999999999999, "true": 0.89}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:dev:7", "state_id": "catalog_lookup_v2:dev:7", "family_id": "catalog_lookup_v2", "split": "dev", "state": "Warehouse audit: item=SKU-00235; color=brown; size=large; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "black, medium", "candidate_1": "cyan, large", "candidate_2": "red, medium", "candidate_3": "brown, large", "candidate_4": "green, small", "candidate_5": "orange, medium", "candidate_6": "gray, large", "candidate_7": "orange, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_3", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 1, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00235\",\"brown\",\"large\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_5": 0, "candidate_6": 0, "candidate_3": 1, "candidate_0": 0, "candidate_7": 0, "candidate_2": 0, "candidate_4": 0, "candidate_1": 0}, "ready": {"false": 0.89, "true": 0.11}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:dev:4", "state_id": "catalog_lookup_v2:dev:4", "family_id": "catalog_lookup_v2", "split": "dev", "state": "Warehouse audit: item=SKU-00248; color=gray; size=medium; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "cyan, medium", "candidate_1": "gray, medium", "candidate_2": "red, medium", "candidate_3": "brown, large", "candidate_4": "gray, large", "candidate_5": "purple, small", "candidate_6": "pink, medium", "candidate_7": "red, large", "candidate_8": "black, small", "candidate_9": "gray, small", "candidate_10": "cyan, large", "candidate_11": "cyan, small", "candidate_12": "orange, medium", "candidate_13": "green, medium", "candidate_14": "blue, small", "candidate_15": "red, small", "candidate_16": "yellow, small", "candidate_17": "black, medium", "candidate_18": "green, large", "candidate_19": "brown, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00248\",\"gray\",\"medium\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_14": 0, "candidate_9": 0, "candidate_0": 0, "candidate_13": 0, "candidate_2": 0, "candidate_7": 0, "candidate_18": 0, "candidate_12": 0, "candidate_11": 0, "candidate_16": 0, "candidate_10": 0, "candidate_5": 0, "candidate_15": 0, "candidate_19": 0, "candidate_1": 1, "candidate_4": 0, "candidate_6": 0, "candidate_17": 0, "candidate_8": 0, "candidate_3": 0}, "ready": {"false": 0.74, "true": 0.26}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:dev:13", "state_id": "catalog_lookup_v2:dev:13", "family_id": "catalog_lookup_v2", "split": "dev", "state": "Warehouse audit: item=SKU-00232; color=brown; size=large; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, medium", "candidate_1": "black, large", "candidate_2": "red, large", "candidate_3": "yellow, large", "candidate_4": "brown, small", "candidate_5": "pink, medium", "candidate_6": "cyan, small", "candidate_7": "yellow, small", "candidate_8": "brown, large", "candidate_9": "pink, large", "candidate_10": "orange, medium", "candidate_11": "gray, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_8", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 1, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00232\",\"brown\",\"large\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_3": 0, "candidate_2": 0, "candidate_10": 0, "candidate_0": 0, "candidate_11": 0, "candidate_1": 0, "candidate_8": 1, "candidate_4": 0, "candidate_6": 0, "candidate_5": 0, "candidate_7": 0, "candidate_9": 0}, "ready": {"false": 0.72, "true": 0.28}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:dev:14", "state_id": "catalog_lookup_v2:dev:14", "family_id": "catalog_lookup_v2", "split": "dev", "state": "Warehouse audit: item=SKU-00245; color=gray; size=small; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "white, small", "candidate_1": "orange, large", "candidate_2": "pink, small", "candidate_3": "purple, small", "candidate_4": "pink, large", "candidate_5": "red, large", "candidate_6": "cyan, medium", "candidate_7": "cyan, small", "candidate_8": "brown, small", "candidate_9": "green, medium", "candidate_10": "gray, medium", "candidate_11": "pink, medium", "candidate_12": "black, small", "candidate_13": "brown, large", "candidate_14": "red, medium", "candidate_15": "orange, medium", "candidate_16": "gray, small", "candidate_17": "green, small", "candidate_18": "blue, large", "candidate_19": "white, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_16", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 1, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00245\",\"gray\",\"small\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_15": 0, "candidate_19": 0, "candidate_4": 0, "candidate_10": 0, "candidate_9": 0, "candidate_18": 0, "candidate_6": 0, "candidate_13": 0, "candidate_7": 0, "candidate_0": 0, "candidate_8": 0, "candidate_11": 0, "candidate_3": 0, "candidate_2": 0, "candidate_14": 0, "candidate_12": 0, "candidate_16": 1, "candidate_17": 0, "candidate_1": 0, "candidate_5": 0}, "ready": {"false": 0.10999999999999999, "true": 0.89}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:dev:9", "state_id": "catalog_lookup_v2:dev:9", "family_id": "catalog_lookup_v2", "split": "dev", "state": "Warehouse audit: item=SKU-00183; color=orange; size=medium; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "gray, small", "candidate_1": "brown, large", "candidate_2": "orange, small", "candidate_3": "red, large", "candidate_4": "black, large", "candidate_5": "cyan, large", "candidate_6": "purple, medium", "candidate_7": "red, small", "candidate_8": "green, small", "candidate_9": "purple, small", "candidate_10": "orange, medium", "candidate_11": "gray, large", "candidate_12": "white, small", "candidate_13": "green, medium", "candidate_14": "yellow, small", "candidate_15": "pink, large", "candidate_16": "blue, small", "candidate_17": "blue, large", "candidate_18": "orange, large", "candidate_19": "gray, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_10", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 1, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00183\",\"orange\",\"medium\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 0, "candidate_3": 0, "candidate_18": 0, "candidate_6": 0, "candidate_13": 0, "candidate_17": 0, "candidate_4": 0, "candidate_11": 0, "candidate_16": 0, "candidate_19": 0, "candidate_14": 0, "candidate_8": 0, "candidate_9": 0, "candidate_15": 0, "candidate_5": 0, "candidate_10": 1, "candidate_7": 0, "candidate_2": 0, "candidate_12": 0, "candidate_1": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:dev:15", "state_id": "catalog_lookup_v2:dev:15", "family_id": "catalog_lookup_v2", "split": "dev", "state": "Warehouse audit: item=SKU-00013; color=red; size=medium; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "orange, small", "candidate_1": "red, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00013\",\"red\",\"medium\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 1, "candidate_0": 0}, "ready": {"false": 0.10999999999999999, "true": 0.89}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:dev:18", "state_id": "catalog_lookup_v2:dev:18", "family_id": "catalog_lookup_v2", "split": "dev", "state": "Warehouse audit: item=SKU-00147; color=black; size=small; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "yellow, large", "candidate_1": "green, small", "candidate_2": "gray, large", "candidate_3": "white, small", "candidate_4": "gray, medium", "candidate_5": "blue, medium", "candidate_6": "orange, medium", "candidate_7": "white, medium", "candidate_8": "orange, small", "candidate_9": "cyan, large", "candidate_10": "black, large", "candidate_11": "black, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_11", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00147\",\"black\",\"small\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_8": 0, "candidate_3": 0, "candidate_4": 0, "candidate_9": 0, "candidate_10": 0, "candidate_7": 0, "candidate_1": 0, "candidate_5": 0, "candidate_6": 0, "candidate_0": 0, "candidate_11": 1, "candidate_2": 0}, "ready": {"false": 0.84, "true": 0.16}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:dev:19", "state_id": "catalog_lookup_v2:dev:19", "family_id": "catalog_lookup_v2", "split": "dev", "state": "Warehouse audit: item=SKU-00014; color=red; size=medium; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "blue, medium", "candidate_1": "blue, small", "candidate_2": "black, medium", "candidate_3": "white, medium", "candidate_4": "red, large", "candidate_5": "gray, medium", "candidate_6": "red, small", "candidate_7": "brown, medium", "candidate_8": "gray, small", "candidate_9": "purple, large", "candidate_10": "cyan, medium", "candidate_11": "cyan, large", "candidate_12": "yellow, medium", "candidate_13": "orange, large", "candidate_14": "black, large", "candidate_15": "white, small", "candidate_16": "red, medium", "candidate_17": "white, large", "candidate_18": "purple, small", "candidate_19": "gray, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_16", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 1, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00014\",\"red\",\"medium\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_16": 1, "candidate_8": 0, "candidate_4": 0, "candidate_11": 0, "candidate_0": 0, "candidate_1": 0, "candidate_6": 0, "candidate_10": 0, "candidate_13": 0, "candidate_5": 0, "candidate_9": 0, "candidate_2": 0, "candidate_3": 0, "candidate_7": 0, "candidate_12": 0, "candidate_14": 0, "candidate_15": 0, "candidate_17": 0, "candidate_19": 0, "candidate_18": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:dev:17", "state_id": "catalog_lookup_v2:dev:17", "family_id": "catalog_lookup_v2", "split": "dev", "state": "Warehouse audit: item=SKU-00188; color=orange; size=large; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "orange, large", "candidate_1": "gray, small", "candidate_2": "yellow, small", "candidate_3": "pink, large", "candidate_4": "purple, small", "candidate_5": "black, small", "candidate_6": "cyan, medium", "candidate_7": "green, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00188\",\"orange\",\"large\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_3": 0, "candidate_1": 0, "candidate_6": 0, "candidate_5": 0, "candidate_4": 0, "candidate_0": 1, "candidate_7": 0, "candidate_2": 0}, "ready": {"false": 0.18000000000000005, "true": 0.82}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:dev:16", "state_id": "catalog_lookup_v2:dev:16", "family_id": "catalog_lookup_v2", "split": "dev", "state": "Warehouse audit: item=SKU-00263; color=gray; size=large; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "gray, large", "candidate_1": "brown, small", "candidate_2": "yellow, large", "candidate_3": "cyan, large", "candidate_4": "white, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00263\",\"gray\",\"large\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_3": 0, "candidate_4": 0, "candidate_1": 0, "candidate_2": 0, "candidate_0": 1}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:dev:20", "state_id": "catalog_lookup_v2:dev:20", "family_id": "catalog_lookup_v2", "split": "dev", "state": "Warehouse audit: item=SKU-00280; color=cyan; size=large; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "cyan, large", "candidate_1": "brown, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00280\",\"cyan\",\"large\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 0.73, "true": 0.27}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:dev:22", "state_id": "catalog_lookup_v2:dev:22", "family_id": "catalog_lookup_v2", "split": "dev", "state": "Warehouse audit: item=SKU-00062; color=green; size=medium; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "orange, medium", "candidate_1": "brown, large", "candidate_2": "yellow, small", "candidate_3": "black, medium", "candidate_4": "green, medium", "candidate_5": "gray, small", "candidate_6": "red, medium", "candidate_7": "pink, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_4", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 1, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00062\",\"green\",\"medium\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 0, "candidate_3": 0, "candidate_4": 1, "candidate_6": 0, "candidate_1": 0, "candidate_5": 0, "candidate_2": 0, "candidate_7": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:dev:21", "state_id": "catalog_lookup_v2:dev:21", "family_id": "catalog_lookup_v2", "split": "dev", "state": "Warehouse audit: item=SKU-00125; color=white; size=small; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "red, medium", "candidate_1": "gray, small", "candidate_2": "yellow, large", "candidate_3": "white, small", "candidate_4": "yellow, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_3", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 1, "candidate_4": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00125\",\"white\",\"small\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_4": 0, "candidate_0": 0, "candidate_3": 1, "candidate_2": 0}, "ready": {"false": 0.12, "true": 0.88}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:dev:2", "state_id": "known_chance_v2:dev:2", "family_id": "known_chance_v2", "split": "dev", "state": "Sampling experiment: exactly one ball is drawn uniformly from a bag. Counts: red=7, brown=7, purple=1, green=5, cyan=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "purple": "The sampled ball is purple", "red": "The sampled ball is red", "brown": "The sampled ball is brown", "green": "The sampled ball is green"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a red ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: red, green", "The drawn color is one of: brown, cyan", "The drawn color is one of: purple"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.23076923076923078, "purple": 0.038461538461538464, "red": 0.2692307692307692, "brown": 0.2692307692307692, "green": 0.19230769230769232}, "focus": {"false": 0.7307692307692308, "true": 0.2692307692307692}, "group": {"0": 0.46153846153846156, "1": 0.5, "2": 0.038461538461538464}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"7/26\"],[\"cyan\",\"3/13\"],[\"green\",\"5/26\"],[\"purple\",\"1/26\"],[\"red\",\"7/26\"]]", "template_id": "known_chance_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0.05, "purple": 0.06, "green": 0.04, "red": 0.76, "brown": 0.09}, "focus": {"false": 0.78, "true": 0.22}, "group": {"0": 0.65, "1": 0.31, "2": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:dev:23", "state_id": "catalog_lookup_v2:dev:23", "family_id": "catalog_lookup_v2", "split": "dev", "state": "Warehouse audit: item=SKU-00026; color=blue; size=small; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "brown, medium", "candidate_1": "green, medium", "candidate_2": "blue, medium", "candidate_3": "orange, small", "candidate_4": "blue, small", "candidate_5": "cyan, medium", "candidate_6": "red, small", "candidate_7": "purple, small", "candidate_8": "purple, medium", "candidate_9": "black, medium", "candidate_10": "black, small", "candidate_11": "yellow, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_4", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 1, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00026\",\"blue\",\"small\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_4": 1, "candidate_1": 0, "candidate_6": 0, "candidate_9": 0, "candidate_0": 0, "candidate_2": 0, "candidate_8": 0, "candidate_7": 0, "candidate_10": 0, "candidate_11": 0, "candidate_3": 0, "candidate_5": 0}, "ready": {"false": 0.85, "true": 0.15}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:dev:4", "state_id": "known_chance_v2:dev:4", "family_id": "known_chance_v2", "split": "dev", "state": "Sampling experiment: exactly one ball is drawn uniformly from a bag. Counts: black=7, red=4, orange=3, green=4, gray=3, brown=5, pink=1, cyan=4, white=4, yellow=4, blue=3, purple=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"orange": "The sampled ball is orange", "blue": "The sampled ball is blue", "pink": "The sampled ball is pink", "purple": "The sampled ball is purple", "green": "The sampled ball is green", "cyan": "The sampled ball is cyan", "white": "The sampled ball is white", "black": "The sampled ball is black", "red": "The sampled ball is red", "brown": "The sampled ball is brown", "gray": "The sampled ball is gray", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a orange ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: black, green, pink, yellow", "The drawn color is one of: red, gray, cyan, blue", "The drawn color is one of: orange, brown, white, purple"]}}, "gold": {}, "gold_probs": {"outcome": {"orange": 0.06, "blue": 0.06, "pink": 0.02, "purple": 0.16, "green": 0.08, "cyan": 0.08, "white": 0.08, "black": 0.14, "red": 0.08, "brown": 0.1, "gray": 0.06, "yellow": 0.08}, "focus": {"false": 0.94, "true": 0.06}, "group": {"0": 0.32, "1": 0.28, "2": 0.4}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"7/50\"],[\"blue\",\"3/50\"],[\"brown\",\"1/10\"],[\"cyan\",\"2/25\"],[\"gray\",\"3/50\"],[\"green\",\"2/25\"],[\"orange\",\"3/50\"],[\"pink\",\"1/50\"],[\"purple\",\"4/25\"],[\"red\",\"2/25\"],[\"white\",\"2/25\"],[\"yellow\",\"2/25\"]]", "template_id": "known_chance_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0, "yellow": 0, "gray": 0, "red": 0, "purple": 0.84, "cyan": 0, "pink": 0.03, "black": 0.13, "blue": 0, "white": 0, "brown": 0, "orange": 0}, "focus": {"false": 0.88, "true": 0.12}, "group": {"0": 0.13, "1": 0.1, "2": 0.77}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:dev:0", "state_id": "known_chance_v2:dev:0", "family_id": "known_chance_v2", "split": "dev", "state": "Sampling experiment: exactly one ball is drawn uniformly from a bag. Counts: blue=5, white=7, brown=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"brown": "The sampled ball is brown", "blue": "The sampled ball is blue", "white": "The sampled ball is white"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue", "The drawn color is one of: white", "The drawn color is one of: brown"]}}, "gold": {}, "gold_probs": {"outcome": {"brown": 0.2, "blue": 0.3333333333333333, "white": 0.4666666666666667}, "focus": {"false": 0.6666666666666667, "true": 0.3333333333333333}, "group": {"0": 0.3333333333333333, "1": 0.4666666666666667, "2": 0.2}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"1/3\"],[\"brown\",\"1/5\"],[\"white\",\"7/15\"]]", "template_id": "known_chance_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"brown": 0.02, "blue": 0.34, "white": 0.64}, "focus": {"false": 0.76, "true": 0.24}, "group": {"0": 0.05, "1": 0.9299999999999999, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:dev:1", "state_id": "known_chance_v2:dev:1", "family_id": "known_chance_v2", "split": "dev", "state": "Sampling experiment: exactly one ball is drawn uniformly from a bag. Counts: orange=7, blue=2, green=7, white=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "orange": "The sampled ball is orange", "white": "The sampled ball is white", "green": "The sampled ball is green"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a white ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: orange, white", "The drawn color is one of: blue", "The drawn color is one of: green"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.11764705882352941, "orange": 0.4117647058823529, "white": 0.058823529411764705, "green": 0.4117647058823529}, "focus": {"false": 0.9411764705882353, "true": 0.058823529411764705}, "group": {"0": 0.47058823529411764, "1": 0.11764705882352941, "2": 0.4117647058823529}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"2/17\"],[\"green\",\"7/17\"],[\"orange\",\"7/17\"],[\"white\",\"1/17\"]]", "template_id": "known_chance_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0.07, "orange": 0.91, "white": 0.01, "blue": 0.01}, "focus": {"false": 0.9299999999999999, "true": 0.07}, "group": {"0": 0.67, "1": 0.01, "2": 0.32}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:dev:3", "state_id": "known_chance_v2:dev:3", "family_id": "known_chance_v2", "split": "dev", "state": "Sampling experiment: exactly one ball is drawn uniformly from a bag. Counts: brown=7, pink=7, purple=9, cyan=7, green=6, black=2, yellow=8, orange=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"yellow": "The sampled ball is yellow", "black": "The sampled ball is black", "cyan": "The sampled ball is cyan", "orange": "The sampled ball is orange", "brown": "The sampled ball is brown", "green": "The sampled ball is green", "purple": "The sampled ball is purple", "pink": "The sampled ball is pink"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: brown, cyan, yellow", "The drawn color is one of: pink, green, orange", "The drawn color is one of: purple, black"]}}, "gold": {}, "gold_probs": {"outcome": {"yellow": 0.16666666666666666, "black": 0.041666666666666664, "cyan": 0.14583333333333334, "orange": 0.041666666666666664, "brown": 0.14583333333333334, "green": 0.125, "purple": 0.1875, "pink": 0.14583333333333334}, "focus": {"false": 0.9583333333333334, "true": 0.041666666666666664}, "group": {"0": 0.45833333333333337, "1": 0.3125, "2": 0.22916666666666666}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/24\"],[\"brown\",\"7/48\"],[\"cyan\",\"7/48\"],[\"green\",\"1/8\"],[\"orange\",\"1/24\"],[\"pink\",\"7/48\"],[\"purple\",\"3/16\"],[\"yellow\",\"1/6\"]]", "template_id": "known_chance_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.01, "purple": 0.82, "cyan": 0.01, "brown": 0.11, "yellow": 0.02, "black": 0.01, "orange": 0.01, "green": 0.01}, "focus": {"false": 0.92, "true": 0.08}, "group": {"0": 0.72, "1": 0.05, "2": 0.23}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:dev:5", "state_id": "known_chance_v2:dev:5", "family_id": "known_chance_v2", "split": "dev", "state": "Sampling experiment: exactly one ball is drawn uniformly from a bag. Counts: orange=8, purple=5, gray=9. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"orange": "The sampled ball is orange", "gray": "The sampled ball is gray", "purple": "The sampled ball is purple"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a orange ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: orange", "The drawn color is one of: purple", "The drawn color is one of: gray"]}}, "gold": {}, "gold_probs": {"outcome": {"orange": 0.36363636363636365, "gray": 0.4090909090909091, "purple": 0.22727272727272727}, "focus": {"false": 0.6363636363636364, "true": 0.36363636363636365}, "group": {"0": 0.36363636363636365, "1": 0.22727272727272727, "2": 0.4090909090909091}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"gray\",\"9/22\"],[\"orange\",\"4/11\"],[\"purple\",\"5/22\"]]", "template_id": "known_chance_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.02, "orange": 0.32, "gray": 0.66}, "focus": {"false": 0.71, "true": 0.29}, "group": {"0": 0.11, "1": 0.02, "2": 0.87}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:dev:6", "state_id": "known_chance_v2:dev:6", "family_id": "known_chance_v2", "split": "dev", "state": "Sampling experiment: exactly one ball is drawn uniformly from a bag. Counts: brown=4, orange=1, gray=5, blue=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"orange": "The sampled ball is orange", "brown": "The sampled ball is brown", "blue": "The sampled ball is blue", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a orange ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: brown, blue", "The drawn color is one of: orange", "The drawn color is one of: gray"]}}, "gold": {}, "gold_probs": {"outcome": {"orange": 0.07142857142857142, "brown": 0.2857142857142857, "blue": 0.2857142857142857, "gray": 0.35714285714285715}, "focus": {"false": 0.9285714285714286, "true": 0.07142857142857142}, "group": {"0": 0.5714285714285714, "1": 0.07142857142857142, "2": 0.35714285714285715}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"2/7\"],[\"brown\",\"2/7\"],[\"gray\",\"5/14\"],[\"orange\",\"1/14\"]]", "template_id": "known_chance_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"brown": 0.16, "orange": 0.02, "gray": 0.8, "blue": 0.01}, "focus": {"false": 0.88, "true": 0.12}, "group": {"0": 0.21, "1": 0.01, "2": 0.78}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:dev:9", "state_id": "known_chance_v2:dev:9", "family_id": "known_chance_v2", "split": "dev", "state": "Sampling experiment: exactly one ball is drawn uniformly from a bag. Counts: pink=6, white=1, brown=3, purple=3, red=5, blue=7, orange=9, gray=6, black=9, green=2, cyan=1, yellow=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"orange": "The sampled ball is orange", "black": "The sampled ball is black", "red": "The sampled ball is red", "green": "The sampled ball is green", "cyan": "The sampled ball is cyan", "gray": "The sampled ball is gray", "pink": "The sampled ball is pink", "white": "The sampled ball is white", "blue": "The sampled ball is blue", "purple": "The sampled ball is purple", "brown": "The sampled ball is brown", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a pink ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: pink, purple, orange, green", "The drawn color is one of: white, red, gray, cyan", "The drawn color is one of: brown, blue, black, yellow"]}}, "gold": {}, "gold_probs": {"outcome": {"orange": 0.16981132075471697, "black": 0.16981132075471697, "red": 0.09433962264150944, "green": 0.03773584905660377, "cyan": 0.018867924528301886, "gray": 0.11320754716981132, "pink": 0.11320754716981132, "white": 0.018867924528301886, "blue": 0.1320754716981132, "purple": 0.05660377358490566, "brown": 0.05660377358490566, "yellow": 0.018867924528301886}, "focus": {"false": 0.8867924528301887, "true": 0.11320754716981132}, "group": {"0": 0.3773584905660377, "1": 0.24528301886792453, "2": 0.3773584905660377}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"9/53\"],[\"blue\",\"7/53\"],[\"brown\",\"3/53\"],[\"cyan\",\"1/53\"],[\"gray\",\"6/53\"],[\"green\",\"2/53\"],[\"orange\",\"9/53\"],[\"pink\",\"6/53\"],[\"purple\",\"3/53\"],[\"red\",\"5/53\"],[\"white\",\"1/53\"],[\"yellow\",\"1/53\"]]", "template_id": "known_chance_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.01, "cyan": 0.02, "yellow": 0.01, "white": 0.01, "gray": 0.02, "red": 0.01, "blue": 0.11, "orange": 0.53, "pink": 0.11, "green": 0, "black": 0.16, "brown": 0.01}, "focus": {"false": 0.87, "true": 0.13}, "group": {"0": 0.4, "1": 0.03, "2": 0.57}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:dev:13", "state_id": "known_chance_v2:dev:13", "family_id": "known_chance_v2", "split": "dev", "state": "Sampling experiment: exactly one ball is drawn uniformly from a bag. Counts: green=4, pink=2, white=4, black=9, yellow=4, blue=6, orange=2, gray=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "orange": "The sampled ball is orange", "black": "The sampled ball is black", "green": "The sampled ball is green", "yellow": "The sampled ball is yellow", "pink": "The sampled ball is pink", "white": "The sampled ball is white", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: green, black, orange", "The drawn color is one of: pink, yellow, gray", "The drawn color is one of: white, blue"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.18181818181818182, "orange": 0.06060606060606061, "black": 0.2727272727272727, "green": 0.12121212121212122, "yellow": 0.12121212121212122, "pink": 0.06060606060606061, "white": 0.12121212121212122, "gray": 0.06060606060606061}, "focus": {"false": 0.7272727272727273, "true": 0.2727272727272727}, "group": {"0": 0.45454545454545453, "1": 0.24242424242424243, "2": 0.30303030303030304}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"3/11\"],[\"blue\",\"2/11\"],[\"gray\",\"2/33\"],[\"green\",\"4/33\"],[\"orange\",\"2/33\"],[\"pink\",\"2/33\"],[\"white\",\"4/33\"],[\"yellow\",\"4/33\"]]", "template_id": "known_chance_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0, "pink": 0, "yellow": 0, "green": 0.03, "gray": 0, "black": 0.96, "orange": 0, "blue": 0.01}, "focus": {"false": 0.74, "true": 0.26}, "group": {"0": 0.8, "1": 0.02, "2": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:dev:7", "state_id": "known_chance_v2:dev:7", "family_id": "known_chance_v2", "split": "dev", "state": "Sampling experiment: exactly one ball is drawn uniformly from a bag. Counts: red=2, gray=9, purple=7, black=8, brown=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "purple": "The sampled ball is purple", "gray": "The sampled ball is gray", "red": "The sampled ball is red", "brown": "The sampled ball is brown"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a gray ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: red, black", "The drawn color is one of: gray, brown", "The drawn color is one of: purple"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.27586206896551724, "purple": 0.2413793103448276, "gray": 0.3103448275862069, "red": 0.06896551724137931, "brown": 0.10344827586206896}, "focus": {"false": 0.6896551724137931, "true": 0.3103448275862069}, "group": {"0": 0.3448275862068966, "1": 0.41379310344827586, "2": 0.2413793103448276}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"8/29\"],[\"brown\",\"3/29\"],[\"gray\",\"9/29\"],[\"purple\",\"7/29\"],[\"red\",\"2/29\"]]", "template_id": "known_chance_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"gray": 0.9, "purple": 0.01, "red": 0.07, "brown": 0, "black": 0.02}, "focus": {"false": 0.72, "true": 0.28}, "group": {"0": 0.1, "1": 0.83, "2": 0.07}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:dev:10", "state_id": "known_chance_v2:dev:10", "family_id": "known_chance_v2", "split": "dev", "state": "Sampling experiment: exactly one ball is drawn uniformly from a bag. Counts: blue=4, black=2, gray=9. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "gray": "The sampled ball is gray", "blue": "The sampled ball is blue"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue", "The drawn color is one of: black", "The drawn color is one of: gray"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.13333333333333333, "gray": 0.6, "blue": 0.26666666666666666}, "focus": {"false": 0.8666666666666667, "true": 0.13333333333333333}, "group": {"0": 0.26666666666666666, "1": 0.13333333333333333, "2": 0.6}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"2/15\"],[\"blue\",\"4/15\"],[\"gray\",\"3/5\"]]", "template_id": "known_chance_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"black": 0.01, "blue": 0.04, "gray": 0.95}, "focus": {"false": 0.8200000000000001, "true": 0.18}, "group": {"0": 0.01, "1": 0.01, "2": 0.98}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:dev:12", "state_id": "known_chance_v2:dev:12", "family_id": "known_chance_v2", "split": "dev", "state": "Sampling experiment: exactly one ball is drawn uniformly from a bag. Counts: blue=1, cyan=3, pink=4, red=1, black=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"red": "The sampled ball is red", "pink": "The sampled ball is pink", "cyan": "The sampled ball is cyan", "black": "The sampled ball is black", "blue": "The sampled ball is blue"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a red ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue, red", "The drawn color is one of: cyan, black", "The drawn color is one of: pink"]}}, "gold": {}, "gold_probs": {"outcome": {"red": 0.08333333333333333, "pink": 0.3333333333333333, "cyan": 0.25, "black": 0.25, "blue": 0.08333333333333333}, "focus": {"false": 0.9166666666666666, "true": 0.08333333333333333}, "group": {"0": 0.16666666666666666, "1": 0.5, "2": 0.3333333333333333}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/4\"],[\"blue\",\"1/12\"],[\"cyan\",\"1/4\"],[\"pink\",\"1/3\"],[\"red\",\"1/12\"]]", "template_id": "known_chance_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.92, "black": 0.01, "cyan": 0.03, "red": 0, "blue": 0.04}, "focus": {"false": 0.86, "true": 0.14}, "group": {"0": 0.01, "1": 0.32, "2": 0.67}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:dev:8", "state_id": "known_chance_v2:dev:8", "family_id": "known_chance_v2", "split": "dev", "state": "Sampling experiment: exactly one ball is drawn uniformly from a bag. Counts: orange=2, green=8, yellow=9, purple=4, cyan=5, gray=1, red=7, brown=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "orange": "The sampled ball is orange", "yellow": "The sampled ball is yellow", "purple": "The sampled ball is purple", "red": "The sampled ball is red", "gray": "The sampled ball is gray", "brown": "The sampled ball is brown", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a red ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: orange, purple, red", "The drawn color is one of: green, cyan, brown", "The drawn color is one of: yellow, gray"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.2, "orange": 0.05, "yellow": 0.225, "purple": 0.1, "red": 0.175, "gray": 0.025, "brown": 0.1, "cyan": 0.125}, "focus": {"false": 0.825, "true": 0.175}, "group": {"0": 0.325, "1": 0.42500000000000004, "2": 0.25}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"1/10\"],[\"cyan\",\"1/8\"],[\"gray\",\"1/40\"],[\"green\",\"1/5\"],[\"orange\",\"1/20\"],[\"purple\",\"1/10\"],[\"red\",\"7/40\"],[\"yellow\",\"9/40\"]]", "template_id": "known_chance_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"brown": 0, "green": 0.02, "purple": 0, "cyan": 0, "red": 0.01, "orange": 0.01, "gray": 0.01, "yellow": 0.9400000000000001}, "focus": {"false": 0.85, "true": 0.15}, "group": {"0": 0.02, "1": 0.54, "2": 0.44}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:dev:11", "state_id": "known_chance_v2:dev:11", "family_id": "known_chance_v2", "split": "dev", "state": "Sampling experiment: exactly one ball is drawn uniformly from a bag. Counts: cyan=7, yellow=5, green=2, gray=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"yellow": "The sampled ball is yellow", "cyan": "The sampled ball is cyan", "gray": "The sampled ball is gray", "green": "The sampled ball is green"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a green ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: cyan, gray", "The drawn color is one of: yellow", "The drawn color is one of: green"]}}, "gold": {}, "gold_probs": {"outcome": {"yellow": 0.2777777777777778, "cyan": 0.3888888888888889, "gray": 0.2222222222222222, "green": 0.1111111111111111}, "focus": {"false": 0.8888888888888888, "true": 0.1111111111111111}, "group": {"0": 0.6111111111111112, "1": 0.2777777777777778, "2": 0.1111111111111111}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"7/18\"],[\"gray\",\"2/9\"],[\"green\",\"1/9\"],[\"yellow\",\"5/18\"]]", "template_id": "known_chance_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"gray": 0, "green": 0.01, "yellow": 0.01, "cyan": 0.98}, "focus": {"false": 0.88, "true": 0.12}, "group": {"0": 0.97, "1": 0.02, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:dev:16", "state_id": "known_chance_v2:dev:16", "family_id": "known_chance_v2", "split": "dev", "state": "Sampling experiment: exactly one ball is drawn uniformly from a bag. Counts: gray=5, purple=7, orange=2, brown=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"gray": "The sampled ball is gray", "orange": "The sampled ball is orange", "purple": "The sampled ball is purple", "brown": "The sampled ball is brown"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a brown ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: gray, brown", "The drawn color is one of: purple", "The drawn color is one of: orange"]}}, "gold": {}, "gold_probs": {"outcome": {"gray": 0.29411764705882354, "orange": 0.11764705882352941, "purple": 0.4117647058823529, "brown": 0.17647058823529413}, "focus": {"false": 0.8235294117647058, "true": 0.17647058823529413}, "group": {"0": 0.47058823529411764, "1": 0.4117647058823529, "2": 0.11764705882352941}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"3/17\"],[\"gray\",\"5/17\"],[\"orange\",\"2/17\"],[\"purple\",\"7/17\"]]", "template_id": "known_chance_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.77, "orange": 0.01, "brown": 0.01, "gray": 0.21}, "focus": {"false": 0.83, "true": 0.17}, "group": {"0": 0.24, "1": 0.75, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:dev:17", "state_id": "known_chance_v2:dev:17", "family_id": "known_chance_v2", "split": "dev", "state": "Sampling experiment: exactly one ball is drawn uniformly from a bag. Counts: yellow=9, black=5, brown=8, pink=4, purple=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "pink": "The sampled ball is pink", "brown": "The sampled ball is brown", "purple": "The sampled ball is purple", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: yellow, pink", "The drawn color is one of: black, purple", "The drawn color is one of: brown"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.17857142857142858, "pink": 0.14285714285714285, "brown": 0.2857142857142857, "purple": 0.07142857142857142, "yellow": 0.32142857142857145}, "focus": {"false": 0.6785714285714286, "true": 0.32142857142857145}, "group": {"0": 0.4642857142857143, "1": 0.25, "2": 0.2857142857142857}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"5/28\"],[\"brown\",\"2/7\"],[\"pink\",\"1/7\"],[\"purple\",\"1/14\"],[\"yellow\",\"9/28\"]]", "template_id": "known_chance_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.01, "yellow": 0.98, "black": 0, "brown": 0.01, "pink": 0}, "focus": {"false": 0.73, "true": 0.27}, "group": {"0": 0.82, "1": 0.04, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:dev:18", "state_id": "known_chance_v2:dev:18", "family_id": "known_chance_v2", "split": "dev", "state": "Sampling experiment: exactly one ball is drawn uniformly from a bag. Counts: black=1, white=5, blue=3, brown=5, gray=5, pink=8, red=3, cyan=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"pink": "The sampled ball is pink", "black": "The sampled ball is black", "red": "The sampled ball is red", "white": "The sampled ball is white", "cyan": "The sampled ball is cyan", "brown": "The sampled ball is brown", "blue": "The sampled ball is blue", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: black, brown, red", "The drawn color is one of: white, gray, cyan", "The drawn color is one of: blue, pink"]}}, "gold": {}, "gold_probs": {"outcome": {"pink": 0.21052631578947367, "black": 0.02631578947368421, "red": 0.07894736842105263, "white": 0.13157894736842105, "cyan": 0.21052631578947367, "brown": 0.13157894736842105, "blue": 0.07894736842105263, "gray": 0.13157894736842105}, "focus": {"false": 0.9210526315789473, "true": 0.07894736842105263}, "group": {"0": 0.23684210526315788, "1": 0.47368421052631576, "2": 0.2894736842105263}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/38\"],[\"blue\",\"3/38\"],[\"brown\",\"5/38\"],[\"cyan\",\"4/19\"],[\"gray\",\"5/38\"],[\"pink\",\"4/19\"],[\"red\",\"3/38\"],[\"white\",\"5/38\"]]", "template_id": "known_chance_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.42, "white": 0.04, "cyan": 0.18, "brown": 0.02, "blue": 0.01, "gray": 0.01, "red": 0.01, "black": 0.31}, "focus": {"false": 0.87, "true": 0.13}, "group": {"0": 0.03, "1": 0.58, "2": 0.39}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:dev:14", "state_id": "known_chance_v2:dev:14", "family_id": "known_chance_v2", "split": "dev", "state": "Sampling experiment: exactly one ball is drawn uniformly from a bag. Counts: red=4, pink=5, blue=1, white=3, cyan=6, orange=5, black=1, green=9, purple=5, gray=8, yellow=4, brown=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "pink": "The sampled ball is pink", "purple": "The sampled ball is purple", "white": "The sampled ball is white", "blue": "The sampled ball is blue", "cyan": "The sampled ball is cyan", "black": "The sampled ball is black", "red": "The sampled ball is red", "orange": "The sampled ball is orange", "brown": "The sampled ball is brown", "gray": "The sampled ball is gray", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a white ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: red, white, black, gray", "The drawn color is one of: pink, cyan, green, yellow", "The drawn color is one of: blue, orange, purple, brown"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.16071428571428573, "pink": 0.08928571428571429, "purple": 0.08928571428571429, "white": 0.05357142857142857, "blue": 0.017857142857142856, "cyan": 0.10714285714285714, "black": 0.017857142857142856, "red": 0.07142857142857142, "orange": 0.08928571428571429, "brown": 0.08928571428571429, "gray": 0.14285714285714285, "yellow": 0.07142857142857142}, "focus": {"false": 0.9464285714285714, "true": 0.05357142857142857}, "group": {"0": 0.2857142857142857, "1": 0.4285714285714286, "2": 0.2857142857142857}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/56\"],[\"blue\",\"1/56\"],[\"brown\",\"5/56\"],[\"cyan\",\"3/28\"],[\"gray\",\"1/7\"],[\"green\",\"9/56\"],[\"orange\",\"5/56\"],[\"pink\",\"5/56\"],[\"purple\",\"5/56\"],[\"red\",\"1/14\"],[\"white\",\"3/56\"],[\"yellow\",\"1/14\"]]", "template_id": "known_chance_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0, "purple": 0, "brown": 0, "orange": 0, "black": 0, "pink": 0.01, "yellow": 0, "gray": 0.02, "cyan": 0.02, "blue": 0.02, "green": 0.9, "red": 0.03}, "focus": {"false": 0.9, "true": 0.1}, "group": {"0": 0.03, "1": 0.96, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:dev:19", "state_id": "known_chance_v2:dev:19", "family_id": "known_chance_v2", "split": "dev", "state": "Sampling experiment: exactly one ball is drawn uniformly from a bag. Counts: blue=5, cyan=9, orange=8, green=7, brown=7, purple=2, black=2, white=6, gray=4, yellow=1, red=9, pink=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "white": "The sampled ball is white", "green": "The sampled ball is green", "purple": "The sampled ball is purple", "cyan": "The sampled ball is cyan", "pink": "The sampled ball is pink", "orange": "The sampled ball is orange", "blue": "The sampled ball is blue", "yellow": "The sampled ball is yellow", "gray": "The sampled ball is gray", "red": "The sampled ball is red", "brown": "The sampled ball is brown"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a green ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue, green, black, yellow", "The drawn color is one of: cyan, brown, white, red", "The drawn color is one of: orange, purple, gray, pink"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.030303030303030304, "white": 0.09090909090909091, "green": 0.10606060606060606, "purple": 0.030303030303030304, "cyan": 0.13636363636363635, "pink": 0.09090909090909091, "orange": 0.12121212121212122, "blue": 0.07575757575757576, "yellow": 0.015151515151515152, "gray": 0.06060606060606061, "red": 0.13636363636363635, "brown": 0.10606060606060606}, "focus": {"false": 0.8939393939393939, "true": 0.10606060606060606}, "group": {"0": 0.2272727272727273, "1": 0.46969696969696967, "2": 0.30303030303030304}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/33\"],[\"blue\",\"5/66\"],[\"brown\",\"7/66\"],[\"cyan\",\"3/22\"],[\"gray\",\"2/33\"],[\"green\",\"7/66\"],[\"orange\",\"4/33\"],[\"pink\",\"1/11\"],[\"purple\",\"1/33\"],[\"red\",\"3/22\"],[\"white\",\"1/11\"],[\"yellow\",\"1/66\"]]", "template_id": "known_chance_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"blue": 0.12, "red": 0.22, "gray": 0.01, "orange": 0.03, "green": 0.01, "black": 0.01, "white": 0.01, "yellow": 0.1, "pink": 0.01, "purple": 0, "cyan": 0.47, "brown": 0.01}, "focus": {"false": 0.87, "true": 0.13}, "group": {"0": 0.01, "1": 0.98, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:dev:21", "state_id": "known_chance_v2:dev:21", "family_id": "known_chance_v2", "split": "dev", "state": "Sampling experiment: exactly one ball is drawn uniformly from a bag. Counts: black=5, cyan=7, brown=4, blue=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "black": "The sampled ball is black", "cyan": "The sampled ball is cyan", "brown": "The sampled ball is brown"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: black, blue", "The drawn color is one of: cyan", "The drawn color is one of: brown"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.30434782608695654, "black": 0.21739130434782608, "cyan": 0.30434782608695654, "brown": 0.17391304347826086}, "focus": {"false": 0.6956521739130435, "true": 0.30434782608695654}, "group": {"0": 0.5217391304347826, "1": 0.30434782608695654, "2": 0.17391304347826086}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"5/23\"],[\"blue\",\"7/23\"],[\"brown\",\"4/23\"],[\"cyan\",\"7/23\"]]", "template_id": "known_chance_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0.4, "blue": 0.13, "brown": 0.03, "black": 0.44}, "focus": {"false": 0.74, "true": 0.26}, "group": {"0": 0.34, "1": 0.64, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:dev:20", "state_id": "known_chance_v2:dev:20", "family_id": "known_chance_v2", "split": "dev", "state": "Sampling experiment: exactly one ball is drawn uniformly from a bag. Counts: black=5, brown=5, red=9. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"brown": "The sampled ball is brown", "black": "The sampled ball is black", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: black", "The drawn color is one of: brown", "The drawn color is one of: red"]}}, "gold": {}, "gold_probs": {"outcome": {"brown": 0.2631578947368421, "black": 0.2631578947368421, "red": 0.47368421052631576}, "focus": {"false": 0.736842105263158, "true": 0.2631578947368421}, "group": {"0": 0.2631578947368421, "1": 0.2631578947368421, "2": 0.47368421052631576}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"5/19\"],[\"brown\",\"5/19\"],[\"red\",\"9/19\"]]", "template_id": "known_chance_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"black": 0.16, "brown": 0.02, "red": 0.82}, "focus": {"false": 0.77, "true": 0.23}, "group": {"0": 0.03, "1": 0.01, "2": 0.96}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:dev:15", "state_id": "known_chance_v2:dev:15", "family_id": "known_chance_v2", "split": "dev", "state": "Sampling experiment: exactly one ball is drawn uniformly from a bag. Counts: black=2, gray=3, white=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"white": "The sampled ball is white", "black": "The sampled ball is black", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a white ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: black", "The drawn color is one of: gray", "The drawn color is one of: white"]}}, "gold": {}, "gold_probs": {"outcome": {"white": 0.4444444444444444, "black": 0.2222222222222222, "gray": 0.3333333333333333}, "focus": {"false": 0.5555555555555556, "true": 0.4444444444444444}, "group": {"0": 0.2222222222222222, "1": 0.3333333333333333, "2": 0.4444444444444444}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"2/9\"],[\"gray\",\"1/3\"],[\"white\",\"4/9\"]]", "template_id": "known_chance_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"gray": 0.03, "black": 0.18, "white": 0.79}, "focus": {"false": 0.6699999999999999, "true": 0.33}, "group": {"0": 0.04, "1": 0.03, "2": 0.9299999999999999}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:dev:22", "state_id": "known_chance_v2:dev:22", "family_id": "known_chance_v2", "split": "dev", "state": "Sampling experiment: exactly one ball is drawn uniformly from a bag. Counts: purple=5, orange=5, red=9, white=6, black=9. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"purple": "The sampled ball is purple", "white": "The sampled ball is white", "orange": "The sampled ball is orange", "red": "The sampled ball is red", "black": "The sampled ball is black"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a red ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: purple, white", "The drawn color is one of: orange, black", "The drawn color is one of: red"]}}, "gold": {}, "gold_probs": {"outcome": {"purple": 0.14705882352941177, "white": 0.17647058823529413, "orange": 0.14705882352941177, "red": 0.2647058823529412, "black": 0.2647058823529412}, "focus": {"false": 0.7352941176470589, "true": 0.2647058823529412}, "group": {"0": 0.3235294117647059, "1": 0.4117647058823529, "2": 0.2647058823529412}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"9/34\"],[\"orange\",\"5/34\"],[\"purple\",\"5/34\"],[\"red\",\"9/34\"],[\"white\",\"3/17\"]]", "template_id": "known_chance_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0.03, "black": 0.11, "purple": 0.15, "orange": 0.02, "red": 0.69}, "focus": {"false": 0.73, "true": 0.27}, "group": {"0": 0.05, "1": 0.28, "2": 0.67}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:dev:23", "state_id": "known_chance_v2:dev:23", "family_id": "known_chance_v2", "split": "dev", "state": "Sampling experiment: exactly one ball is drawn uniformly from a bag. Counts: yellow=4, purple=5, gray=6, cyan=6, pink=2, blue=8, white=4, orange=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "pink": "The sampled ball is pink", "gray": "The sampled ball is gray", "white": "The sampled ball is white", "cyan": "The sampled ball is cyan", "orange": "The sampled ball is orange", "yellow": "The sampled ball is yellow", "purple": "The sampled ball is purple"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a white ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: yellow, cyan, white", "The drawn color is one of: purple, pink, orange", "The drawn color is one of: gray, blue"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.21621621621621623, "pink": 0.05405405405405406, "gray": 0.16216216216216217, "white": 0.10810810810810811, "cyan": 0.16216216216216217, "orange": 0.05405405405405406, "yellow": 0.10810810810810811, "purple": 0.13513513513513514}, "focus": {"false": 0.8918918918918919, "true": 0.10810810810810811}, "group": {"0": 0.3783783783783784, "1": 0.24324324324324326, "2": 0.3783783783783784}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"8/37\"],[\"cyan\",\"6/37\"],[\"gray\",\"6/37\"],[\"orange\",\"2/37\"],[\"pink\",\"2/37\"],[\"purple\",\"5/37\"],[\"white\",\"4/37\"],[\"yellow\",\"4/37\"]]", "template_id": "known_chance_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"gray": 0.02, "yellow": 0.03, "purple": 0.01, "blue": 0.9, "white": 0.01, "cyan": 0.02, "orange": 0.01, "pink": 0}, "focus": {"false": 0.86, "true": 0.14}, "group": {"0": 0.04, "1": 0.01, "2": 0.95}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:dev:24", "state_id": "known_chance_v2:dev:24", "family_id": "known_chance_v2", "split": "dev", "state": "Sampling experiment: exactly one ball is drawn uniformly from a bag. Counts: white=2, yellow=3, purple=9, black=9, gray=2, cyan=9, blue=6, red=1, orange=1, pink=7, green=9, brown=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "cyan": "The sampled ball is cyan", "orange": "The sampled ball is orange", "yellow": "The sampled ball is yellow", "brown": "The sampled ball is brown", "white": "The sampled ball is white", "green": "The sampled ball is green", "gray": "The sampled ball is gray", "red": "The sampled ball is red", "purple": "The sampled ball is purple", "black": "The sampled ball is black", "pink": "The sampled ball is pink"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white, black, blue, pink", "The drawn color is one of: yellow, gray, red, green", "The drawn color is one of: purple, cyan, orange, brown"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.09523809523809523, "cyan": 0.14285714285714285, "orange": 0.015873015873015872, "yellow": 0.047619047619047616, "brown": 0.07936507936507936, "white": 0.031746031746031744, "green": 0.14285714285714285, "gray": 0.031746031746031744, "red": 0.015873015873015872, "purple": 0.14285714285714285, "black": 0.14285714285714285, "pink": 0.1111111111111111}, "focus": {"false": 0.8571428571428572, "true": 0.14285714285714285}, "group": {"0": 0.38095238095238093, "1": 0.23809523809523808, "2": 0.38095238095238093}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/7\"],[\"blue\",\"2/21\"],[\"brown\",\"5/63\"],[\"cyan\",\"1/7\"],[\"gray\",\"2/63\"],[\"green\",\"1/7\"],[\"orange\",\"1/63\"],[\"pink\",\"1/9\"],[\"purple\",\"1/7\"],[\"red\",\"1/63\"],[\"white\",\"2/63\"],[\"yellow\",\"1/21\"]]", "template_id": "known_chance_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0.09, "orange": 0.01, "red": 0.02, "purple": 0.55, "pink": 0.01, "gray": 0, "brown": 0.01, "white": 0.07, "cyan": 0.05, "black": 0.15, "blue": 0.03, "yellow": 0.01}, "focus": {"false": 0.83, "true": 0.17}, "group": {"0": 0.23, "1": 0.07, "2": 0.7}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:dev:25", "state_id": "known_chance_v2:dev:25", "family_id": "known_chance_v2", "split": "dev", "state": "Sampling experiment: exactly one ball is drawn uniformly from a bag. Counts: yellow=1, cyan=8, pink=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"yellow": "The sampled ball is yellow", "pink": "The sampled ball is pink", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: yellow", "The drawn color is one of: cyan", "The drawn color is one of: pink"]}}, "gold": {}, "gold_probs": {"outcome": {"yellow": 0.0625, "pink": 0.4375, "cyan": 0.5}, "focus": {"false": 0.5, "true": 0.5}, "group": {"0": 0.0625, "1": 0.5, "2": 0.4375}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"1/2\"],[\"pink\",\"7/16\"],[\"yellow\",\"1/16\"]]", "template_id": "known_chance_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.02, "yellow": 0.08, "cyan": 0.9}, "focus": {"false": 0.56, "true": 0.44}, "group": {"0": 0.01, "1": 0.96, "2": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:dev:26", "state_id": "known_chance_v2:dev:26", "family_id": "known_chance_v2", "split": "dev", "state": "Sampling experiment: exactly one ball is drawn uniformly from a bag. Counts: green=7, yellow=2, black=7, cyan=9. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "cyan": "The sampled ball is cyan", "green": "The sampled ball is green", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: green, cyan", "The drawn color is one of: yellow", "The drawn color is one of: black"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.28, "cyan": 0.36, "green": 0.28, "yellow": 0.08}, "focus": {"false": 0.64, "true": 0.36}, "group": {"0": 0.64, "1": 0.08, "2": 0.28}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"7/25\"],[\"cyan\",\"9/25\"],[\"green\",\"7/25\"],[\"yellow\",\"2/25\"]]", "template_id": "known_chance_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"yellow": 0.02, "green": 0.15, "cyan": 0.81, "black": 0.02}, "focus": {"false": 0.73, "true": 0.27}, "group": {"0": 0.98, "1": 0.01, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:dev:29", "state_id": "known_chance_v2:dev:29", "family_id": "known_chance_v2", "split": "dev", "state": "Sampling experiment: exactly one ball is drawn uniformly from a bag. Counts: pink=9, black=9, purple=5, red=5, cyan=3, orange=4, green=5, blue=9, brown=9, yellow=5, white=2, gray=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"pink": "The sampled ball is pink", "brown": "The sampled ball is brown", "cyan": "The sampled ball is cyan", "gray": "The sampled ball is gray", "purple": "The sampled ball is purple", "black": "The sampled ball is black", "green": "The sampled ball is green", "yellow": "The sampled ball is yellow", "red": "The sampled ball is red", "white": "The sampled ball is white", "blue": "The sampled ball is blue", "orange": "The sampled ball is orange"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a red ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: pink, red, green, yellow", "The drawn color is one of: black, cyan, blue, white", "The drawn color is one of: purple, orange, brown, gray"]}}, "gold": {}, "gold_probs": {"outcome": {"pink": 0.1323529411764706, "brown": 0.1323529411764706, "cyan": 0.04411764705882353, "gray": 0.04411764705882353, "purple": 0.07352941176470588, "black": 0.1323529411764706, "green": 0.07352941176470588, "yellow": 0.07352941176470588, "red": 0.07352941176470588, "white": 0.029411764705882353, "blue": 0.1323529411764706, "orange": 0.058823529411764705}, "focus": {"false": 0.9264705882352942, "true": 0.07352941176470588}, "group": {"0": 0.35294117647058826, "1": 0.3382352941176471, "2": 0.3088235294117647}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"9/68\"],[\"blue\",\"9/68\"],[\"brown\",\"9/68\"],[\"cyan\",\"3/68\"],[\"gray\",\"3/68\"],[\"green\",\"5/68\"],[\"orange\",\"1/17\"],[\"pink\",\"9/68\"],[\"purple\",\"5/68\"],[\"red\",\"5/68\"],[\"white\",\"1/34\"],[\"yellow\",\"5/68\"]]", "template_id": "known_chance_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"brown": 0.02, "green": 0.01, "white": 0.09, "purple": 0.01, "cyan": 0.01, "red": 0.01, "orange": 0.01, "gray": 0.02, "black": 0.1, "blue": 0.09, "pink": 0.62, "yellow": 0.01}, "focus": {"false": 0.87, "true": 0.13}, "group": {"0": 0.46, "1": 0.42, "2": 0.12}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:dev:27", "state_id": "known_chance_v2:dev:27", "family_id": "known_chance_v2", "split": "dev", "state": "Sampling experiment: exactly one ball is drawn uniformly from a bag. Counts: red=5, green=2, cyan=3, white=8, brown=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"white": "The sampled ball is white", "green": "The sampled ball is green", "brown": "The sampled ball is brown", "cyan": "The sampled ball is cyan", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a green ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: red, white", "The drawn color is one of: green, brown", "The drawn color is one of: cyan"]}}, "gold": {}, "gold_probs": {"outcome": {"white": 0.34782608695652173, "green": 0.08695652173913043, "brown": 0.21739130434782608, "cyan": 0.13043478260869565, "red": 0.21739130434782608}, "focus": {"false": 0.9130434782608696, "true": 0.08695652173913043}, "group": {"0": 0.5652173913043478, "1": 0.30434782608695654, "2": 0.13043478260869565}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"5/23\"],[\"cyan\",\"3/23\"],[\"green\",\"2/23\"],[\"red\",\"5/23\"],[\"white\",\"8/23\"]]", "template_id": "known_chance_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0.9299999999999999, "brown": 0.01, "red": 0.04, "cyan": 0.01, "green": 0.01}, "focus": {"false": 0.86, "true": 0.14}, "group": {"0": 0.98, "1": 0.01, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:dev:28", "state_id": "known_chance_v2:dev:28", "family_id": "known_chance_v2", "split": "dev", "state": "Sampling experiment: exactly one ball is drawn uniformly from a bag. Counts: orange=5, cyan=5, white=3, blue=7, purple=7, black=7, green=7, yellow=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "black": "The sampled ball is black", "orange": "The sampled ball is orange", "white": "The sampled ball is white", "purple": "The sampled ball is purple", "cyan": "The sampled ball is cyan", "yellow": "The sampled ball is yellow", "green": "The sampled ball is green"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a green ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: orange, blue, green", "The drawn color is one of: cyan, purple, yellow", "The drawn color is one of: white, black"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.14893617021276595, "black": 0.14893617021276595, "orange": 0.10638297872340426, "white": 0.06382978723404255, "purple": 0.14893617021276595, "cyan": 0.10638297872340426, "yellow": 0.1276595744680851, "green": 0.14893617021276595}, "focus": {"false": 0.851063829787234, "true": 0.14893617021276595}, "group": {"0": 0.40425531914893614, "1": 0.3829787234042553, "2": 0.21276595744680848}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"7/47\"],[\"blue\",\"7/47\"],[\"cyan\",\"5/47\"],[\"green\",\"7/47\"],[\"orange\",\"5/47\"],[\"purple\",\"7/47\"],[\"white\",\"3/47\"],[\"yellow\",\"6/47\"]]", "template_id": "known_chance_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0.02, "green": 0.05, "blue": 0.42, "purple": 0.06, "white": 0.06, "black": 0.04, "orange": 0.28, "yellow": 0.07}, "focus": {"false": 0.8200000000000001, "true": 0.18}, "group": {"0": 0.84, "1": 0.11, "2": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:dev:31", "state_id": "known_chance_v2:dev:31", "family_id": "known_chance_v2", "split": "dev", "state": "Sampling experiment: exactly one ball is drawn uniformly from a bag. Counts: pink=2, blue=8, yellow=4, purple=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"yellow": "The sampled ball is yellow", "blue": "The sampled ball is blue", "pink": "The sampled ball is pink", "purple": "The sampled ball is purple"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a purple ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: pink, purple", "The drawn color is one of: blue", "The drawn color is one of: yellow"]}}, "gold": {}, "gold_probs": {"outcome": {"yellow": 0.18181818181818182, "blue": 0.36363636363636365, "pink": 0.09090909090909091, "purple": 0.36363636363636365}, "focus": {"false": 0.6363636363636364, "true": 0.36363636363636365}, "group": {"0": 0.4545454545454546, "1": 0.36363636363636365, "2": 0.18181818181818182}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"4/11\"],[\"pink\",\"1/11\"],[\"purple\",\"4/11\"],[\"yellow\",\"2/11\"]]", "template_id": "known_chance_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"blue": 0.8, "purple": 0.05, "pink": 0.13, "yellow": 0.02}, "focus": {"false": 0.74, "true": 0.26}, "group": {"0": 0.17, "1": 0.81, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:dev:30", "state_id": "known_chance_v2:dev:30", "family_id": "known_chance_v2", "split": "dev", "state": "Sampling experiment: exactly one ball is drawn uniformly from a bag. Counts: purple=6, orange=7, yellow=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"yellow": "The sampled ball is yellow", "purple": "The sampled ball is purple", "orange": "The sampled ball is orange"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a orange ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: purple", "The drawn color is one of: orange", "The drawn color is one of: yellow"]}}, "gold": {}, "gold_probs": {"outcome": {"yellow": 0.23529411764705882, "purple": 0.35294117647058826, "orange": 0.4117647058823529}, "focus": {"false": 0.5882352941176471, "true": 0.4117647058823529}, "group": {"0": 0.35294117647058826, "1": 0.4117647058823529, "2": 0.23529411764705882}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"orange\",\"7/17\"],[\"purple\",\"6/17\"],[\"yellow\",\"4/17\"]]", "template_id": "known_chance_v2:dev"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"orange": 0.78, "yellow": 0.01, "purple": 0.21}, "focus": {"false": 0.6599999999999999, "true": 0.34}, "group": {"0": 0.05, "1": 0.9299999999999999, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:calibration:0", "state_id": "smart_home_v2:calibration:0", "family_id": "smart_home_v2", "split": "calibration", "state": "Control audit: off / bedroom / speaker; authorized yes, occupancy 0, hour 9.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom speaker", "unit_1": "office lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bedroom\",\"speaker\",true,0]", "template_id": "smart_home_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0.050000000000000044, "true": 0.95}, "risk": {"0": 0.98, "1": 0.02, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:calibration:2", "state_id": "smart_home_v2:calibration:2", "family_id": "smart_home_v2", "split": "calibration", "state": "Control audit: off / bathroom / heater; authorized yes, occupancy 2, hour 0.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garden heater", "unit_1": "bathroom heater", "unit_2": "library fan", "unit_3": "bathroom fan", "unit_4": "living room lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0, "unit_3": 0, "unit_4": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bathroom\",\"heater\",true,2]", "template_id": "smart_home_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_4": 0, "unit_3": 0, "unit_2": 0, "unit_1": 1}, "execute": {"false": 0.040000000000000036, "true": 0.96}, "risk": {"0": 0.99, "1": 0, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:calibration:3", "state_id": "smart_home_v2:calibration:3", "family_id": "smart_home_v2", "split": "calibration", "state": "Control audit: off / bathroom / speaker; authorized yes, occupancy 2, hour 21.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library lamp", "unit_1": "office fan", "unit_2": "kitchen heater", "unit_3": "attic speaker", "unit_4": "garage fan", "unit_5": "dining room speaker", "unit_6": "bathroom speaker", "unit_7": "living room lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_6", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 1, "unit_7": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bathroom\",\"speaker\",true,2]", "template_id": "smart_home_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_4": 0, "unit_3": 0, "unit_7": 0, "unit_1": 0, "unit_6": 1, "unit_0": 0, "unit_5": 0}, "execute": {"false": 0.050000000000000044, "true": 0.95}, "risk": {"0": 0.97, "1": 0, "2": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:calibration:1", "state_id": "smart_home_v2:calibration:1", "family_id": "smart_home_v2", "split": "calibration", "state": "Control audit: on / office / fan; authorized no, occupancy 3, hour 13.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "office fan", "unit_1": "garden speaker", "unit_2": "library fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"office\",\"fan\",false,3]", "template_id": "smart_home_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_2": 0, "unit_0": 1}, "execute": {"false": 0.95, "true": 0.05}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:calibration:4", "state_id": "smart_home_v2:calibration:4", "family_id": "smart_home_v2", "split": "calibration", "state": "Control audit: on / bedroom / fan; authorized yes, occupancy 1, hour 20.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom fan", "unit_1": "balcony speaker", "unit_2": "kitchen speaker", "unit_3": "bathroom lamp", "unit_4": "library lamp", "unit_5": "office speaker", "unit_6": "kitchen heater", "unit_7": "bathroom heater", "unit_8": "attic fan", "unit_9": "bathroom speaker", "unit_10": "office lamp", "unit_11": "dining room speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bedroom\",\"fan\",true,1]", "template_id": "smart_home_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_3": 0, "unit_8": 0, "unit_11": 0, "unit_2": 0, "unit_9": 0, "unit_4": 0, "unit_10": 0, "unit_6": 0, "unit_0": 1, "unit_7": 0, "unit_5": 0, "unit_1": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:calibration:5", "state_id": "smart_home_v2:calibration:5", "family_id": "smart_home_v2", "split": "calibration", "state": "Control audit: off / attic / speaker; authorized no, occupancy 1, hour 17.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "hall speaker", "unit_1": "attic speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"attic\",\"speaker\",false,1]", "template_id": "smart_home_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 0.9299999999999999, "true": 0.07}, "risk": {"0": 0.02, "1": 0, "2": 0.98}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:calibration:6", "state_id": "smart_home_v2:calibration:6", "family_id": "smart_home_v2", "split": "calibration", "state": "Control audit: on / bedroom / speaker; authorized no, occupancy 1, hour 21.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "office heater", "unit_1": "living room heater", "unit_2": "bedroom speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bedroom\",\"speaker\",false,1]", "template_id": "smart_home_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 1, "unit_1": 0, "unit_0": 0}, "execute": {"false": 0.95, "true": 0.05}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:calibration:8", "state_id": "smart_home_v2:calibration:8", "family_id": "smart_home_v2", "split": "calibration", "state": "Control audit: off / bathroom / fan; authorized no, occupancy 3, hour 20.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room speaker", "unit_1": "dining room fan", "unit_2": "living room fan", "unit_3": "bathroom fan", "unit_4": "balcony fan", "unit_5": "bedroom lamp", "unit_6": "garden fan", "unit_7": "attic heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_3", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 1, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bathroom\",\"fan\",false,3]", "template_id": "smart_home_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_7": 0, "unit_6": 0, "unit_4": 0, "unit_1": 0, "unit_3": 1, "unit_5": 0, "unit_0": 0}, "execute": {"false": 0.96, "true": 0.04}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:calibration:7", "state_id": "smart_home_v2:calibration:7", "family_id": "smart_home_v2", "split": "calibration", "state": "Control audit: on / garage / heater; authorized yes, occupancy 2, hour 8.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "office speaker", "unit_1": "library lamp", "unit_2": "garage heater", "unit_3": "balcony heater", "unit_4": "living room speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1, "unit_3": 0, "unit_4": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garage\",\"heater\",true,2]", "template_id": "smart_home_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_2": 1, "unit_1": 0, "unit_3": 0, "unit_4": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:calibration:10", "state_id": "smart_home_v2:calibration:10", "family_id": "smart_home_v2", "split": "calibration", "state": "Control audit: off / attic / fan; authorized no, occupancy 1, hour 9.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony lamp", "unit_1": "attic fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"attic\",\"fan\",false,1]", "template_id": "smart_home_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 0.95, "true": 0.05}, "risk": {"0": 0.02, "1": 0, "2": 0.98}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:calibration:9", "state_id": "smart_home_v2:calibration:9", "family_id": "smart_home_v2", "split": "calibration", "state": "Control audit: off / garden / heater; authorized yes, occupancy 0, hour 5.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "dining room speaker", "unit_1": "dining room fan", "unit_2": "living room fan", "unit_3": "office heater", "unit_4": "bathroom heater", "unit_5": "garden heater", "unit_6": "garage heater", "unit_7": "library heater", "unit_8": "hall heater", "unit_9": "dining room lamp", "unit_10": "balcony lamp", "unit_11": "garage speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_5", "execute": false, "risk": 1}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 1, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 1, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garden\",\"heater\",true,0]", "template_id": "smart_home_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_8": 0, "unit_4": 0, "unit_3": 0, "unit_6": 0, "unit_1": 0, "unit_11": 0, "unit_10": 0, "unit_5": 1, "unit_7": 0, "unit_2": 0, "unit_9": 0}, "execute": {"false": 0.85, "true": 0.15}, "risk": {"0": 0.06, "1": 0.92, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:calibration:14", "state_id": "smart_home_v2:calibration:14", "family_id": "smart_home_v2", "split": "calibration", "state": "Control audit: off / garden / heater; authorized no, occupancy 0, hour 16.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "dining room heater", "unit_1": "library speaker", "unit_2": "hall heater", "unit_3": "balcony fan", "unit_4": "bedroom lamp", "unit_5": "office heater", "unit_6": "attic lamp", "unit_7": "library heater", "unit_8": "office lamp", "unit_9": "garden heater", "unit_10": "garden fan", "unit_11": "bathroom speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_9", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 1, "unit_10": 0, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garden\",\"heater\",false,0]", "template_id": "smart_home_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_2": 0, "unit_4": 0, "unit_7": 0, "unit_5": 0, "unit_3": 0, "unit_1": 0, "unit_10": 0, "unit_8": 0, "unit_6": 0, "unit_11": 0, "unit_9": 1}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:calibration:11", "state_id": "smart_home_v2:calibration:11", "family_id": "smart_home_v2", "split": "calibration", "state": "Control audit: off / office / fan; authorized no, occupancy 0, hour 8.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony fan", "unit_1": "living room heater", "unit_2": "office fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"office\",\"fan\",false,0]", "template_id": "smart_home_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0.99, "unit_1": 0, "unit_0": 0.01}, "execute": {"false": 0.9299999999999999, "true": 0.07}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:calibration:13", "state_id": "smart_home_v2:calibration:13", "family_id": "smart_home_v2", "split": "calibration", "state": "Control audit: off / balcony / speaker; authorized yes, occupancy 1, hour 20.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library fan", "unit_1": "bedroom fan", "unit_2": "hall speaker", "unit_3": "garden speaker", "unit_4": "garden heater", "unit_5": "hall lamp", "unit_6": "hall fan", "unit_7": "balcony speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_7", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"balcony\",\"speaker\",true,1]", "template_id": "smart_home_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_3": 0, "unit_6": 0, "unit_0": 0, "unit_5": 0, "unit_4": 0, "unit_1": 0, "unit_7": 1}, "execute": {"false": 0.06999999999999995, "true": 0.93}, "risk": {"0": 0.98, "1": 0, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:calibration:12", "state_id": "smart_home_v2:calibration:12", "family_id": "smart_home_v2", "split": "calibration", "state": "Control audit: on / bedroom / heater; authorized no, occupancy 3, hour 20.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "kitchen speaker", "unit_1": "garage speaker", "unit_2": "library lamp", "unit_3": "balcony lamp", "unit_4": "bedroom heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_4", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bedroom\",\"heater\",false,3]", "template_id": "smart_home_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_4": 1, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_0": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:calibration:15", "state_id": "smart_home_v2:calibration:15", "family_id": "smart_home_v2", "split": "calibration", "state": "Control audit: on / dining room / speaker; authorized yes, occupancy 1, hour 14.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "dining room speaker", "unit_1": "bathroom heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"dining room\",\"speaker\",true,1]", "template_id": "smart_home_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0.050000000000000044, "true": 0.95}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:calibration:17", "state_id": "smart_home_v2:calibration:17", "family_id": "smart_home_v2", "split": "calibration", "state": "Control audit: off / garden / speaker; authorized yes, occupancy 0, hour 8.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "attic fan", "unit_1": "dining room heater", "unit_2": "bathroom lamp", "unit_3": "attic lamp", "unit_4": "garden speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_4", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garden\",\"speaker\",true,0]", "template_id": "smart_home_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_3": 0, "unit_1": 0, "unit_4": 1, "unit_2": 0}, "execute": {"false": 0.16000000000000003, "true": 0.84}, "risk": {"0": 0.85, "1": 0.13, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:calibration:16", "state_id": "smart_home_v2:calibration:16", "family_id": "smart_home_v2", "split": "calibration", "state": "Control audit: off / kitchen / lamp; authorized yes, occupancy 3, hour 0.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "office fan", "unit_1": "kitchen lamp", "unit_2": "living room heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"kitchen\",\"lamp\",true,3]", "template_id": "smart_home_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_2": 0, "unit_1": 1}, "execute": {"false": 0.06000000000000005, "true": 0.94}, "risk": {"0": 0.99, "1": 0, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:calibration:18", "state_id": "smart_home_v2:calibration:18", "family_id": "smart_home_v2", "split": "calibration", "state": "Control audit: on / kitchen / fan; authorized yes, occupancy 3, hour 6.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "attic speaker", "unit_1": "garden speaker", "unit_2": "bathroom speaker", "unit_3": "kitchen fan", "unit_4": "office speaker", "unit_5": "hall lamp", "unit_6": "dining room fan", "unit_7": "dining room lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_3", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 1, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"kitchen\",\"fan\",true,3]", "template_id": "smart_home_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_4": 0, "unit_3": 1, "unit_7": 0, "unit_1": 0, "unit_6": 0, "unit_0": 0, "unit_5": 0}, "execute": {"false": 0.040000000000000036, "true": 0.96}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:calibration:19", "state_id": "smart_home_v2:calibration:19", "family_id": "smart_home_v2", "split": "calibration", "state": "Control audit: on / garage / lamp; authorized no, occupancy 2, hour 18.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom fan", "unit_1": "hall lamp", "unit_2": "library speaker", "unit_3": "garage lamp", "unit_4": "living room fan", "unit_5": "dining room fan", "unit_6": "attic heater", "unit_7": "office heater", "unit_8": "living room speaker", "unit_9": "bedroom heater", "unit_10": "library heater", "unit_11": "office speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_3", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 1, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garage\",\"lamp\",false,2]", "template_id": "smart_home_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_3": 1, "unit_8": 0, "unit_1": 0, "unit_2": 0, "unit_7": 0, "unit_11": 0, "unit_0": 0, "unit_9": 0, "unit_10": 0, "unit_6": 0, "unit_5": 0, "unit_4": 0}, "execute": {"false": 0.96, "true": 0.04}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:calibration:20", "state_id": "smart_home_v2:calibration:20", "family_id": "smart_home_v2", "split": "calibration", "state": "Control audit: on / bathroom / lamp; authorized yes, occupancy 3, hour 19.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bathroom lamp", "unit_1": "bathroom heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bathroom\",\"lamp\",true,3]", "template_id": "smart_home_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_0": 1}, "execute": {"false": 0.06000000000000005, "true": 0.94}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:calibration:21", "state_id": "smart_home_v2:calibration:21", "family_id": "smart_home_v2", "split": "calibration", "state": "Control audit: off / bedroom / heater; authorized no, occupancy 0, hour 10.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bathroom speaker", "unit_1": "bedroom heater", "unit_2": "balcony heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bedroom\",\"heater\",false,0]", "template_id": "smart_home_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_0": 0, "unit_1": 1}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:calibration:23", "state_id": "smart_home_v2:calibration:23", "family_id": "smart_home_v2", "split": "calibration", "state": "Control audit: on / office / heater; authorized no, occupancy 0, hour 22.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "office heater", "unit_1": "balcony fan", "unit_2": "hall lamp", "unit_3": "bathroom heater", "unit_4": "garage speaker", "unit_5": "garage heater", "unit_6": "living room fan", "unit_7": "living room lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"office\",\"heater\",false,0]", "template_id": "smart_home_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_5": 0, "unit_4": 0, "unit_0": 1, "unit_3": 0, "unit_2": 0, "unit_1": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:calibration:24", "state_id": "smart_home_v2:calibration:24", "family_id": "smart_home_v2", "split": "calibration", "state": "Control audit: on / attic / speaker; authorized yes, occupancy 0, hour 6.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "hall heater", "unit_1": "kitchen fan", "unit_2": "attic speaker", "unit_3": "dining room fan", "unit_4": "living room speaker", "unit_5": "garden fan", "unit_6": "library heater", "unit_7": "bathroom heater", "unit_8": "kitchen speaker", "unit_9": "garage lamp", "unit_10": "office heater", "unit_11": "hall speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"attic\",\"speaker\",true,0]", "template_id": "smart_home_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_9": 0, "unit_5": 0, "unit_4": 0, "unit_1": 0, "unit_8": 0, "unit_10": 0, "unit_2": 1, "unit_11": 0, "unit_0": 0, "unit_6": 0, "unit_7": 0, "unit_3": 0}, "execute": {"false": 0.10999999999999999, "true": 0.89}, "risk": {"0": 0.83, "1": 0.17, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:calibration:25", "state_id": "smart_home_v2:calibration:25", "family_id": "smart_home_v2", "split": "calibration", "state": "Control audit: on / hall / fan; authorized no, occupancy 1, hour 4.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony lamp", "unit_1": "hall fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"hall\",\"fan\",false,1]", "template_id": "smart_home_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 1, "unit_0": 0}, "execute": {"false": 0.95, "true": 0.05}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:calibration:26", "state_id": "smart_home_v2:calibration:26", "family_id": "smart_home_v2", "split": "calibration", "state": "Control audit: on / kitchen / speaker; authorized yes, occupancy 3, hour 2.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room speaker", "unit_1": "kitchen speaker", "unit_2": "garage lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"kitchen\",\"speaker\",true,3]", "template_id": "smart_home_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0}, "execute": {"false": 0.06000000000000005, "true": 0.94}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:calibration:22", "state_id": "smart_home_v2:calibration:22", "family_id": "smart_home_v2", "split": "calibration", "state": "Control audit: on / attic / lamp; authorized yes, occupancy 3, hour 2.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony speaker", "unit_1": "dining room fan", "unit_2": "garden heater", "unit_3": "attic lamp", "unit_4": "balcony heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_3", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 1, "unit_4": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"attic\",\"lamp\",true,3]", "template_id": "smart_home_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_1": 0, "unit_4": 0, "unit_3": 1, "unit_0": 0}, "execute": {"false": 0.06000000000000005, "true": 0.94}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:calibration:27", "state_id": "smart_home_v2:calibration:27", "family_id": "smart_home_v2", "split": "calibration", "state": "Control audit: on / balcony / fan; authorized yes, occupancy 3, hour 15.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony speaker", "unit_1": "kitchen speaker", "unit_2": "balcony fan", "unit_3": "kitchen lamp", "unit_4": "hall speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1, "unit_3": 0, "unit_4": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"balcony\",\"fan\",true,3]", "template_id": "smart_home_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_4": 0, "unit_1": 0, "unit_3": 0, "unit_0": 0, "unit_2": 1}, "execute": {"false": 0.040000000000000036, "true": 0.96}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:calibration:28", "state_id": "smart_home_v2:calibration:28", "family_id": "smart_home_v2", "split": "calibration", "state": "Control audit: off / bathroom / heater; authorized no, occupancy 1, hour 6.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "dining room speaker", "unit_1": "bathroom lamp", "unit_2": "bedroom lamp", "unit_3": "bathroom heater", "unit_4": "kitchen heater", "unit_5": "kitchen lamp", "unit_6": "balcony speaker", "unit_7": "office fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_3", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 1, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bathroom\",\"heater\",false,1]", "template_id": "smart_home_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_3": 1, "unit_6": 0, "unit_0": 0, "unit_4": 0, "unit_5": 0, "unit_1": 0, "unit_7": 0}, "execute": {"false": 0.96, "true": 0.04}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:calibration:29", "state_id": "smart_home_v2:calibration:29", "family_id": "smart_home_v2", "split": "calibration", "state": "Control audit: off / living room / fan; authorized yes, occupancy 3, hour 0.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom lamp", "unit_1": "library heater", "unit_2": "office fan", "unit_3": "garden speaker", "unit_4": "bathroom heater", "unit_5": "living room lamp", "unit_6": "office speaker", "unit_7": "kitchen lamp", "unit_8": "bathroom speaker", "unit_9": "kitchen speaker", "unit_10": "library lamp", "unit_11": "living room fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_11", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"living room\",\"fan\",true,3]", "template_id": "smart_home_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_10": 0, "unit_3": 0, "unit_2": 0, "unit_4": 0, "unit_0": 0, "unit_7": 0, "unit_9": 0, "unit_11": 1, "unit_6": 0, "unit_5": 0, "unit_8": 0}, "execute": {"false": 0.040000000000000036, "true": 0.96}, "risk": {"0": 0.97, "1": 0, "2": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:calibration:30", "state_id": "smart_home_v2:calibration:30", "family_id": "smart_home_v2", "split": "calibration", "state": "Control audit: off / dining room / lamp; authorized yes, occupancy 3, hour 23.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "dining room speaker", "unit_1": "dining room lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"dining room\",\"lamp\",true,3]", "template_id": "smart_home_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 0.040000000000000036, "true": 0.96}, "risk": {"0": 0.99, "1": 0, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:calibration:0", "state_id": "catalog_lookup_v2:calibration:0", "family_id": "catalog_lookup_v2", "split": "calibration", "state": "Inventory review: item=SKU-00037; color=blue; size=medium; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "yellow, large", "candidate_1": "blue, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00037\",\"blue\",\"medium\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 1, "candidate_0": 0}, "ready": {"false": 0.15000000000000002, "true": 0.85}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:calibration:31", "state_id": "smart_home_v2:calibration:31", "family_id": "smart_home_v2", "split": "calibration", "state": "Control audit: off / dining room / speaker; authorized yes, occupancy 0, hour 21.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom lamp", "unit_1": "bedroom fan", "unit_2": "dining room speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"dining room\",\"speaker\",true,0]", "template_id": "smart_home_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 1, "unit_0": 0, "unit_1": 0}, "execute": {"false": 0.050000000000000044, "true": 0.95}, "risk": {"0": 0.98, "1": 0.02, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:calibration:2", "state_id": "catalog_lookup_v2:calibration:2", "family_id": "catalog_lookup_v2", "split": "calibration", "state": "Inventory review: item=SKU-00187; color=orange; size=large; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "orange, medium", "candidate_1": "yellow, medium", "candidate_2": "white, small", "candidate_3": "red, large", "candidate_4": "orange, large", "candidate_5": "pink, small", "candidate_6": "pink, medium", "candidate_7": "gray, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_4", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 1, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00187\",\"orange\",\"large\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_3": 0, "candidate_0": 0, "candidate_6": 0, "candidate_4": 1, "candidate_5": 0, "candidate_1": 0, "candidate_7": 0, "candidate_2": 0}, "ready": {"false": 0.89, "true": 0.11}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:calibration:1", "state_id": "catalog_lookup_v2:calibration:1", "family_id": "catalog_lookup_v2", "split": "calibration", "state": "Inventory review: item=SKU-00213; color=pink; size=large; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, medium", "candidate_1": "green, medium", "candidate_2": "gray, small", "candidate_3": "yellow, large", "candidate_4": "pink, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_4", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 1}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00213\",\"pink\",\"large\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_4": 1, "candidate_2": 0, "candidate_3": 0, "candidate_0": 0, "candidate_1": 0}, "ready": {"false": 0.14, "true": 0.86}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:calibration:3", "state_id": "catalog_lookup_v2:calibration:3", "family_id": "catalog_lookup_v2", "split": "calibration", "state": "Inventory review: item=SKU-00257; color=gray; size=large; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "red, small", "candidate_1": "gray, large", "candidate_2": "green, medium", "candidate_3": "yellow, small", "candidate_4": "cyan, large", "candidate_5": "red, medium", "candidate_6": "brown, large", "candidate_7": "purple, small", "candidate_8": "orange, medium", "candidate_9": "orange, large", "candidate_10": "gray, medium", "candidate_11": "purple, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00257\",\"gray\",\"large\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 1, "candidate_3": 0, "candidate_10": 0, "candidate_5": 0, "candidate_6": 0, "candidate_8": 0, "candidate_2": 0, "candidate_4": 0, "candidate_0": 0, "candidate_7": 0, "candidate_11": 0, "candidate_9": 0}, "ready": {"false": 0.77, "true": 0.23}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:calibration:4", "state_id": "catalog_lookup_v2:calibration:4", "family_id": "catalog_lookup_v2", "split": "calibration", "state": "Inventory review: item=SKU-00070; color=green; size=large; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "black, large", "candidate_1": "brown, medium", "candidate_2": "pink, large", "candidate_3": "orange, medium", "candidate_4": "green, large", "candidate_5": "white, small", "candidate_6": "black, medium", "candidate_7": "purple, medium", "candidate_8": "pink, medium", "candidate_9": "yellow, medium", "candidate_10": "purple, small", "candidate_11": "purple, large", "candidate_12": "pink, small", "candidate_13": "blue, small", "candidate_14": "cyan, large", "candidate_15": "cyan, small", "candidate_16": "cyan, medium", "candidate_17": "blue, large", "candidate_18": "red, large", "candidate_19": "gray, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_4", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 1, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00070\",\"green\",\"large\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_19": 0, "candidate_13": 0, "candidate_12": 0, "candidate_8": 0, "candidate_17": 0, "candidate_16": 0, "candidate_4": 1, "candidate_10": 0, "candidate_14": 0, "candidate_1": 0, "candidate_11": 0, "candidate_3": 0, "candidate_5": 0, "candidate_0": 0, "candidate_7": 0, "candidate_6": 0, "candidate_15": 0, "candidate_2": 0, "candidate_18": 0, "candidate_9": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:calibration:5", "state_id": "catalog_lookup_v2:calibration:5", "family_id": "catalog_lookup_v2", "split": "calibration", "state": "Inventory review: item=SKU-00110; color=purple; size=medium; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, medium", "candidate_1": "brown, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00110\",\"purple\",\"medium\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:calibration:7", "state_id": "catalog_lookup_v2:calibration:7", "family_id": "catalog_lookup_v2", "split": "calibration", "state": "Inventory review: item=SKU-00174; color=orange; size=small; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "black, medium", "candidate_1": "gray, medium", "candidate_2": "white, medium", "candidate_3": "cyan, medium", "candidate_4": "white, small", "candidate_5": "orange, small", "candidate_6": "green, large", "candidate_7": "blue, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_5", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 1, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00174\",\"orange\",\"small\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_5": 1, "candidate_2": 0, "candidate_0": 0, "candidate_3": 0, "candidate_1": 0, "candidate_6": 0, "candidate_4": 0, "candidate_7": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:calibration:8", "state_id": "catalog_lookup_v2:calibration:8", "family_id": "catalog_lookup_v2", "split": "calibration", "state": "Inventory review: item=SKU-00277; color=cyan; size=medium; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "red, medium", "candidate_1": "yellow, large", "candidate_2": "gray, small", "candidate_3": "white, small", "candidate_4": "purple, large", "candidate_5": "cyan, medium", "candidate_6": "cyan, large", "candidate_7": "green, small", "candidate_8": "green, medium", "candidate_9": "brown, medium", "candidate_10": "gray, large", "candidate_11": "orange, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_5", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 1, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00277\",\"cyan\",\"medium\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_10": 0, "candidate_2": 0, "candidate_11": 0, "candidate_4": 0, "candidate_5": 1, "candidate_9": 0, "candidate_8": 0, "candidate_6": 0, "candidate_0": 0, "candidate_1": 0, "candidate_3": 0, "candidate_7": 0}, "ready": {"false": 0.14, "true": 0.86}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:calibration:6", "state_id": "catalog_lookup_v2:calibration:6", "family_id": "catalog_lookup_v2", "split": "calibration", "state": "Inventory review: item=SKU-00243; color=gray; size=small; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "gray, small", "candidate_1": "black, small", "candidate_2": "purple, large", "candidate_3": "green, small", "candidate_4": "black, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00243\",\"gray\",\"small\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_2": 0, "candidate_4": 0, "candidate_1": 0, "candidate_3": 0, "candidate_0": 1}, "ready": {"false": 0.89, "true": 0.11}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:calibration:10", "state_id": "catalog_lookup_v2:calibration:10", "family_id": "catalog_lookup_v2", "split": "calibration", "state": "Inventory review: item=SKU-00112; color=purple; size=large; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, large", "candidate_1": "red, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00112\",\"purple\",\"large\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_0": 1}, "ready": {"false": 0.77, "true": 0.23}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:calibration:9", "state_id": "catalog_lookup_v2:calibration:9", "family_id": "catalog_lookup_v2", "split": "calibration", "state": "Inventory review: item=SKU-00040; color=blue; size=large; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "green, medium", "candidate_1": "green, small", "candidate_2": "blue, medium", "candidate_3": "pink, large", "candidate_4": "red, large", "candidate_5": "purple, small", "candidate_6": "orange, large", "candidate_7": "brown, small", "candidate_8": "blue, large", "candidate_9": "red, small", "candidate_10": "yellow, small", "candidate_11": "cyan, large", "candidate_12": "cyan, medium", "candidate_13": "gray, large", "candidate_14": "pink, small", "candidate_15": "brown, medium", "candidate_16": "purple, large", "candidate_17": "black, large", "candidate_18": "green, large", "candidate_19": "white, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_8", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 1, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00040\",\"blue\",\"large\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_19": 0, "candidate_13": 0, "candidate_3": 0, "candidate_8": 1, "candidate_17": 0, "candidate_16": 0, "candidate_4": 0, "candidate_10": 0, "candidate_14": 0, "candidate_1": 0, "candidate_11": 0, "candidate_12": 0, "candidate_5": 0, "candidate_0": 0, "candidate_7": 0, "candidate_6": 0, "candidate_15": 0, "candidate_2": 0, "candidate_18": 0, "candidate_9": 0}, "ready": {"false": 0.77, "true": 0.23}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:calibration:12", "state_id": "catalog_lookup_v2:calibration:12", "family_id": "catalog_lookup_v2", "split": "calibration", "state": "Inventory review: item=SKU-00195; color=pink; size=small; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, small", "candidate_1": "gray, medium", "candidate_2": "green, large", "candidate_3": "pink, small", "candidate_4": "green, small", "candidate_5": "yellow, medium", "candidate_6": "red, medium", "candidate_7": "gray, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_3", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 1, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00195\",\"pink\",\"small\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_5": 0, "candidate_2": 0, "candidate_6": 0, "candidate_4": 0, "candidate_0": 0, "candidate_7": 0, "candidate_3": 1}, "ready": {"false": 0.88, "true": 0.12}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:calibration:11", "state_id": "catalog_lookup_v2:calibration:11", "family_id": "catalog_lookup_v2", "split": "calibration", "state": "Inventory review: item=SKU-00200; color=pink; size=medium; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "gray, medium", "candidate_1": "pink, medium", "candidate_2": "blue, large", "candidate_3": "purple, small", "candidate_4": "green, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00200\",\"pink\",\"medium\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_3": 0, "candidate_2": 0, "candidate_4": 0, "candidate_1": 1, "candidate_0": 0}, "ready": {"false": 0.74, "true": 0.26}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:calibration:14", "state_id": "catalog_lookup_v2:calibration:14", "family_id": "catalog_lookup_v2", "split": "calibration", "state": "Inventory review: item=SKU-00247; color=gray; size=small; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "gray, large", "candidate_1": "pink, large", "candidate_2": "black, small", "candidate_3": "pink, small", "candidate_4": "blue, large", "candidate_5": "green, medium", "candidate_6": "red, large", "candidate_7": "purple, medium", "candidate_8": "white, large", "candidate_9": "orange, small", "candidate_10": "brown, medium", "candidate_11": "purple, small", "candidate_12": "green, large", "candidate_13": "pink, medium", "candidate_14": "white, small", "candidate_15": "gray, small", "candidate_16": "white, medium", "candidate_17": "yellow, large", "candidate_18": "black, large", "candidate_19": "yellow, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_15", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 1, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00247\",\"gray\",\"small\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_18": 0, "candidate_5": 0, "candidate_12": 0, "candidate_15": 1, "candidate_7": 0, "candidate_4": 0, "candidate_3": 0, "candidate_6": 0, "candidate_14": 0, "candidate_13": 0, "candidate_2": 0, "candidate_11": 0, "candidate_17": 0, "candidate_1": 0, "candidate_16": 0, "candidate_0": 0, "candidate_10": 0, "candidate_8": 0, "candidate_19": 0, "candidate_9": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:calibration:16", "state_id": "catalog_lookup_v2:calibration:16", "family_id": "catalog_lookup_v2", "split": "calibration", "state": "Inventory review: item=SKU-00282; color=cyan; size=large; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "red, large", "candidate_1": "cyan, large", "candidate_2": "yellow, small", "candidate_3": "white, small", "candidate_4": "blue, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00282\",\"cyan\",\"large\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 0, "candidate_3": 0, "candidate_4": 0, "candidate_1": 1, "candidate_2": 0}, "ready": {"false": 0.9, "true": 0.1}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:calibration:15", "state_id": "catalog_lookup_v2:calibration:15", "family_id": "catalog_lookup_v2", "split": "calibration", "state": "Inventory review: item=SKU-00090; color=yellow; size=large; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "yellow, large", "candidate_1": "yellow, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00090\",\"yellow\",\"large\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 0.87, "true": 0.13}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:calibration:13", "state_id": "catalog_lookup_v2:calibration:13", "family_id": "catalog_lookup_v2", "split": "calibration", "state": "Inventory review: item=SKU-00275; color=cyan; size=medium; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "orange, large", "candidate_1": "cyan, medium", "candidate_2": "yellow, large", "candidate_3": "green, medium", "candidate_4": "pink, small", "candidate_5": "gray, small", "candidate_6": "black, large", "candidate_7": "blue, medium", "candidate_8": "pink, medium", "candidate_9": "cyan, large", "candidate_10": "purple, small", "candidate_11": "gray, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00275\",\"cyan\",\"medium\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_9": 0, "candidate_3": 0, "candidate_0": 0, "candidate_6": 0, "candidate_4": 0, "candidate_5": 0, "candidate_1": 1, "candidate_7": 0, "candidate_10": 0, "candidate_11": 0, "candidate_8": 0, "candidate_2": 0}, "ready": {"false": 0.89, "true": 0.11}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:calibration:18", "state_id": "catalog_lookup_v2:calibration:18", "family_id": "catalog_lookup_v2", "split": "calibration", "state": "Inventory review: item=SKU-00114; color=purple; size=large; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "orange, large", "candidate_1": "red, large", "candidate_2": "gray, small", "candidate_3": "pink, large", "candidate_4": "black, small", "candidate_5": "purple, large", "candidate_6": "cyan, small", "candidate_7": "blue, large", "candidate_8": "blue, medium", "candidate_9": "black, large", "candidate_10": "black, medium", "candidate_11": "purple, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_5", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 1, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00114\",\"purple\",\"large\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_9": 0, "candidate_5": 1, "candidate_4": 0, "candidate_0": 0, "candidate_8": 0, "candidate_3": 0, "candidate_1": 0, "candidate_7": 0, "candidate_11": 0, "candidate_6": 0, "candidate_10": 0, "candidate_2": 0}, "ready": {"false": 0.89, "true": 0.11}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:calibration:17", "state_id": "catalog_lookup_v2:calibration:17", "family_id": "catalog_lookup_v2", "split": "calibration", "state": "Inventory review: item=SKU-00246; color=gray; size=small; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "blue, medium", "candidate_1": "red, small", "candidate_2": "cyan, medium", "candidate_3": "brown, small", "candidate_4": "orange, small", "candidate_5": "gray, small", "candidate_6": "pink, medium", "candidate_7": "red, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_5", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 1, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00246\",\"gray\",\"small\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_7": 0, "candidate_5": 1, "candidate_3": 0, "candidate_4": 0, "candidate_6": 0, "candidate_2": 0, "candidate_0": 0, "candidate_1": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:calibration:19", "state_id": "catalog_lookup_v2:calibration:19", "family_id": "catalog_lookup_v2", "split": "calibration", "state": "Inventory review: item=SKU-00132; color=white; size=medium; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "black, large", "candidate_1": "brown, large", "candidate_2": "purple, large", "candidate_3": "cyan, large", "candidate_4": "gray, small", "candidate_5": "yellow, small", "candidate_6": "green, small", "candidate_7": "brown, medium", "candidate_8": "white, small", "candidate_9": "orange, small", "candidate_10": "blue, large", "candidate_11": "white, medium", "candidate_12": "orange, medium", "candidate_13": "pink, medium", "candidate_14": "black, small", "candidate_15": "purple, medium", "candidate_16": "cyan, medium", "candidate_17": "brown, small", "candidate_18": "yellow, large", "candidate_19": "pink, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_11", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 1, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00132\",\"white\",\"medium\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_16": 0, "candidate_2": 0, "candidate_17": 0, "candidate_9": 0, "candidate_0": 0, "candidate_11": 1, "candidate_13": 0, "candidate_10": 0, "candidate_19": 0, "candidate_3": 0, "candidate_4": 0, "candidate_15": 0, "candidate_14": 0, "candidate_7": 0, "candidate_1": 0, "candidate_12": 0, "candidate_6": 0, "candidate_8": 0, "candidate_5": 0, "candidate_18": 0}, "ready": {"false": 0.19999999999999996, "true": 0.8}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:calibration:20", "state_id": "catalog_lookup_v2:calibration:20", "family_id": "catalog_lookup_v2", "split": "calibration", "state": "Inventory review: item=SKU-00196; color=pink; size=small; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "pink, small", "candidate_1": "pink, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00196\",\"pink\",\"small\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_0": 1}, "ready": {"false": 0.18000000000000005, "true": 0.82}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:calibration:21", "state_id": "catalog_lookup_v2:calibration:21", "family_id": "catalog_lookup_v2", "split": "calibration", "state": "Inventory review: item=SKU-00058; color=green; size=medium; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "green, medium", "candidate_1": "green, large", "candidate_2": "gray, large", "candidate_3": "blue, large", "candidate_4": "cyan, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00058\",\"green\",\"medium\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_2": 0, "candidate_4": 0, "candidate_1": 0, "candidate_3": 0, "candidate_0": 1}, "ready": {"false": 0.87, "true": 0.13}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:calibration:22", "state_id": "catalog_lookup_v2:calibration:22", "family_id": "catalog_lookup_v2", "split": "calibration", "state": "Inventory review: item=SKU-00127; color=white; size=small; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "gray, large", "candidate_1": "white, small", "candidate_2": "red, large", "candidate_3": "green, large", "candidate_4": "black, small", "candidate_5": "brown, small", "candidate_6": "gray, medium", "candidate_7": "pink, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00127\",\"white\",\"small\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_4": 0, "candidate_1": 1, "candidate_6": 0, "candidate_0": 0, "candidate_2": 0, "candidate_7": 0, "candidate_3": 0, "candidate_5": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:calibration:0", "state_id": "known_chance_v2:calibration:0", "family_id": "known_chance_v2", "split": "calibration", "state": "Probability review: exactly one ball is drawn uniformly from a bag. Counts: purple=4, red=6, black=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"purple": "The sampled ball is purple", "black": "The sampled ball is black", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a purple ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: purple", "The drawn color is one of: red", "The drawn color is one of: black"]}}, "gold": {}, "gold_probs": {"outcome": {"purple": 0.26666666666666666, "black": 0.3333333333333333, "red": 0.4}, "focus": {"false": 0.7333333333333334, "true": 0.26666666666666666}, "group": {"0": 0.26666666666666666, "1": 0.4, "2": 0.3333333333333333}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/3\"],[\"purple\",\"4/15\"],[\"red\",\"2/5\"]]", "template_id": "known_chance_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"red": 0.72, "purple": 0.25, "black": 0.03}, "focus": {"false": 0.8, "true": 0.2}, "group": {"0": 0.04, "1": 0.92, "2": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:calibration:23", "state_id": "catalog_lookup_v2:calibration:23", "family_id": "catalog_lookup_v2", "split": "calibration", "state": "Inventory review: item=SKU-00116; color=purple; size=large; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "cyan, large", "candidate_1": "cyan, medium", "candidate_2": "brown, small", "candidate_3": "yellow, small", "candidate_4": "pink, small", "candidate_5": "white, medium", "candidate_6": "pink, large", "candidate_7": "purple, large", "candidate_8": "orange, small", "candidate_9": "blue, large", "candidate_10": "yellow, medium", "candidate_11": "green, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_7", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 1, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00116\",\"purple\",\"large\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_5": 0, "candidate_1": 0, "candidate_8": 0, "candidate_3": 0, "candidate_2": 0, "candidate_6": 0, "candidate_7": 1, "candidate_4": 0, "candidate_11": 0, "candidate_0": 0, "candidate_10": 0, "candidate_9": 0}, "ready": {"false": 0.18000000000000005, "true": 0.82}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:calibration:2", "state_id": "known_chance_v2:calibration:2", "family_id": "known_chance_v2", "split": "calibration", "state": "Probability review: exactly one ball is drawn uniformly from a bag. Counts: brown=2, green=7, red=8, purple=7, blue=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "red": "The sampled ball is red", "green": "The sampled ball is green", "purple": "The sampled ball is purple", "brown": "The sampled ball is brown"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: brown, purple", "The drawn color is one of: green, blue", "The drawn color is one of: red"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.25, "red": 0.25, "green": 0.21875, "purple": 0.21875, "brown": 0.0625}, "focus": {"false": 0.75, "true": 0.25}, "group": {"0": 0.28125, "1": 0.46875, "2": 0.25}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"1/4\"],[\"brown\",\"1/16\"],[\"green\",\"7/32\"],[\"purple\",\"7/32\"],[\"red\",\"1/4\"]]", "template_id": "known_chance_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.02, "brown": 0.26, "green": 0.03, "blue": 0.17, "red": 0.52}, "focus": {"false": 0.8, "true": 0.2}, "group": {"0": 0.04, "1": 0.59, "2": 0.37}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:calibration:3", "state_id": "known_chance_v2:calibration:3", "family_id": "known_chance_v2", "split": "calibration", "state": "Probability review: exactly one ball is drawn uniformly from a bag. Counts: orange=4, gray=1, white=4, green=1, cyan=1, purple=4, red=1, pink=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"pink": "The sampled ball is pink", "orange": "The sampled ball is orange", "red": "The sampled ball is red", "gray": "The sampled ball is gray", "white": "The sampled ball is white", "purple": "The sampled ball is purple", "green": "The sampled ball is green", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: orange, green, red", "The drawn color is one of: gray, cyan, pink", "The drawn color is one of: white, purple"]}}, "gold": {}, "gold_probs": {"outcome": {"pink": 0.2727272727272727, "orange": 0.18181818181818182, "red": 0.045454545454545456, "gray": 0.045454545454545456, "white": 0.18181818181818182, "purple": 0.18181818181818182, "green": 0.045454545454545456, "cyan": 0.045454545454545456}, "focus": {"false": 0.9545454545454546, "true": 0.045454545454545456}, "group": {"0": 0.2727272727272727, "1": 0.36363636363636365, "2": 0.36363636363636365}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"1/22\"],[\"gray\",\"1/22\"],[\"green\",\"1/22\"],[\"orange\",\"2/11\"],[\"pink\",\"3/11\"],[\"purple\",\"2/11\"],[\"red\",\"1/22\"],[\"white\",\"2/11\"]]", "template_id": "known_chance_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.92, "gray": 0, "green": 0, "white": 0.01, "orange": 0.06, "purple": 0.01, "red": 0, "cyan": 0}, "focus": {"false": 0.91, "true": 0.09}, "group": {"0": 0.08, "1": 0.65, "2": 0.27}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:calibration:1", "state_id": "known_chance_v2:calibration:1", "family_id": "known_chance_v2", "split": "calibration", "state": "Probability review: exactly one ball is drawn uniformly from a bag. Counts: gray=2, green=5, black=3, yellow=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"gray": "The sampled ball is gray", "black": "The sampled ball is black", "yellow": "The sampled ball is yellow", "green": "The sampled ball is green"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a green ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: gray, yellow", "The drawn color is one of: green", "The drawn color is one of: black"]}}, "gold": {}, "gold_probs": {"outcome": {"gray": 0.125, "black": 0.1875, "yellow": 0.375, "green": 0.3125}, "focus": {"false": 0.6875, "true": 0.3125}, "group": {"0": 0.5, "1": 0.3125, "2": 0.1875}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"3/16\"],[\"gray\",\"1/8\"],[\"green\",\"5/16\"],[\"yellow\",\"3/8\"]]", "template_id": "known_chance_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"yellow": 0.76, "black": 0.01, "green": 0.1, "gray": 0.13}, "focus": {"false": 0.74, "true": 0.26}, "group": {"0": 0.72, "1": 0.25, "2": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:calibration:4", "state_id": "known_chance_v2:calibration:4", "family_id": "known_chance_v2", "split": "calibration", "state": "Probability review: exactly one ball is drawn uniformly from a bag. Counts: blue=5, orange=7, red=8, brown=1, black=1, cyan=6, green=7, pink=3, white=9, yellow=4, gray=1, purple=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "yellow": "The sampled ball is yellow", "cyan": "The sampled ball is cyan", "purple": "The sampled ball is purple", "red": "The sampled ball is red", "pink": "The sampled ball is pink", "blue": "The sampled ball is blue", "white": "The sampled ball is white", "orange": "The sampled ball is orange", "black": "The sampled ball is black", "gray": "The sampled ball is gray", "brown": "The sampled ball is brown"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue, brown, green, yellow", "The drawn color is one of: orange, black, pink, gray", "The drawn color is one of: red, cyan, white, purple"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.1206896551724138, "yellow": 0.06896551724137931, "cyan": 0.10344827586206896, "purple": 0.10344827586206896, "red": 0.13793103448275862, "pink": 0.05172413793103448, "blue": 0.08620689655172414, "white": 0.15517241379310345, "orange": 0.1206896551724138, "black": 0.017241379310344827, "gray": 0.017241379310344827, "brown": 0.017241379310344827}, "focus": {"false": 0.9310344827586207, "true": 0.06896551724137931}, "group": {"0": 0.2931034482758621, "1": 0.20689655172413793, "2": 0.5}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/58\"],[\"blue\",\"5/58\"],[\"brown\",\"1/58\"],[\"cyan\",\"3/29\"],[\"gray\",\"1/58\"],[\"green\",\"7/58\"],[\"orange\",\"7/58\"],[\"pink\",\"3/58\"],[\"purple\",\"3/29\"],[\"red\",\"4/29\"],[\"white\",\"9/58\"],[\"yellow\",\"2/29\"]]", "template_id": "known_chance_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0, "cyan": 0, "brown": 0.01, "pink": 0, "gray": 0.01, "blue": 0.05, "yellow": 0, "white": 0.81, "red": 0.1, "orange": 0.01, "green": 0.01, "black": 0}, "focus": {"false": 0.87, "true": 0.13}, "group": {"0": 0.01, "1": 0.01, "2": 0.98}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:calibration:5", "state_id": "known_chance_v2:calibration:5", "family_id": "known_chance_v2", "split": "calibration", "state": "Probability review: exactly one ball is drawn uniformly from a bag. Counts: green=6, orange=6, red=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"red": "The sampled ball is red", "orange": "The sampled ball is orange", "green": "The sampled ball is green"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a green ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: green", "The drawn color is one of: orange", "The drawn color is one of: red"]}}, "gold": {}, "gold_probs": {"outcome": {"red": 0.25, "orange": 0.375, "green": 0.375}, "focus": {"false": 0.625, "true": 0.375}, "group": {"0": 0.375, "1": 0.375, "2": 0.25}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"green\",\"3/8\"],[\"orange\",\"3/8\"],[\"red\",\"1/4\"]]", "template_id": "known_chance_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0.88, "orange": 0.09, "red": 0.03}, "focus": {"false": 0.6699999999999999, "true": 0.33}, "group": {"0": 0.84, "1": 0.1, "2": 0.06}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:calibration:7", "state_id": "known_chance_v2:calibration:7", "family_id": "known_chance_v2", "split": "calibration", "state": "Probability review: exactly one ball is drawn uniformly from a bag. Counts: brown=4, gray=7, blue=8, orange=8, purple=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"gray": "The sampled ball is gray", "purple": "The sampled ball is purple", "brown": "The sampled ball is brown", "orange": "The sampled ball is orange", "blue": "The sampled ball is blue"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a purple ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: brown, orange", "The drawn color is one of: gray, purple", "The drawn color is one of: blue"]}}, "gold": {}, "gold_probs": {"outcome": {"gray": 0.2413793103448276, "purple": 0.06896551724137931, "brown": 0.13793103448275862, "orange": 0.27586206896551724, "blue": 0.27586206896551724}, "focus": {"false": 0.9310344827586207, "true": 0.06896551724137931}, "group": {"0": 0.41379310344827586, "1": 0.3103448275862069, "2": 0.27586206896551724}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"8/29\"],[\"brown\",\"4/29\"],[\"gray\",\"7/29\"],[\"orange\",\"8/29\"],[\"purple\",\"2/29\"]]", "template_id": "known_chance_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"orange": 0.06, "brown": 0.2, "blue": 0.52, "purple": 0.03, "gray": 0.19}, "focus": {"false": 0.9, "true": 0.1}, "group": {"0": 0.41, "1": 0.2, "2": 0.39}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:calibration:6", "state_id": "known_chance_v2:calibration:6", "family_id": "known_chance_v2", "split": "calibration", "state": "Probability review: exactly one ball is drawn uniformly from a bag. Counts: white=9, blue=1, black=8, gray=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "black": "The sampled ball is black", "white": "The sampled ball is white", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white, gray", "The drawn color is one of: blue", "The drawn color is one of: black"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.043478260869565216, "black": 0.34782608695652173, "white": 0.391304347826087, "gray": 0.21739130434782608}, "focus": {"false": 0.6521739130434783, "true": 0.34782608695652173}, "group": {"0": 0.6086956521739131, "1": 0.043478260869565216, "2": 0.34782608695652173}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"8/23\"],[\"blue\",\"1/23\"],[\"gray\",\"5/23\"],[\"white\",\"9/23\"]]", "template_id": "known_chance_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"black": 0.04, "blue": 0.01, "white": 0.94, "gray": 0.01}, "focus": {"false": 0.76, "true": 0.24}, "group": {"0": 0.85, "1": 0.01, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:calibration:9", "state_id": "known_chance_v2:calibration:9", "family_id": "known_chance_v2", "split": "calibration", "state": "Probability review: exactly one ball is drawn uniformly from a bag. Counts: red=6, pink=6, purple=8, green=1, gray=6, blue=2, brown=4, orange=4, black=1, white=5, yellow=8, cyan=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "cyan": "The sampled ball is cyan", "purple": "The sampled ball is purple", "orange": "The sampled ball is orange", "brown": "The sampled ball is brown", "blue": "The sampled ball is blue", "gray": "The sampled ball is gray", "white": "The sampled ball is white", "black": "The sampled ball is black", "pink": "The sampled ball is pink", "red": "The sampled ball is red", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a pink ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: red, green, brown, white", "The drawn color is one of: pink, gray, orange, yellow", "The drawn color is one of: purple, blue, black, cyan"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.018518518518518517, "cyan": 0.05555555555555555, "purple": 0.14814814814814814, "orange": 0.07407407407407407, "brown": 0.07407407407407407, "blue": 0.037037037037037035, "gray": 0.1111111111111111, "white": 0.09259259259259259, "black": 0.018518518518518517, "pink": 0.1111111111111111, "red": 0.1111111111111111, "yellow": 0.14814814814814814}, "focus": {"false": 0.8888888888888888, "true": 0.1111111111111111}, "group": {"0": 0.2962962962962963, "1": 0.4444444444444444, "2": 0.25925925925925924}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/54\"],[\"blue\",\"1/27\"],[\"brown\",\"2/27\"],[\"cyan\",\"1/18\"],[\"gray\",\"1/9\"],[\"green\",\"1/54\"],[\"orange\",\"2/27\"],[\"pink\",\"1/9\"],[\"purple\",\"4/27\"],[\"red\",\"1/9\"],[\"white\",\"5/54\"],[\"yellow\",\"4/27\"]]", "template_id": "known_chance_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.32, "cyan": 0.01, "yellow": 0.17, "white": 0.01, "gray": 0.03, "pink": 0.03, "red": 0.39, "orange": 0.01, "blue": 0, "black": 0.01, "green": 0.01, "brown": 0.01}, "focus": {"false": 0.85, "true": 0.15}, "group": {"0": 0.08, "1": 0.71, "2": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:calibration:8", "state_id": "known_chance_v2:calibration:8", "family_id": "known_chance_v2", "split": "calibration", "state": "Probability review: exactly one ball is drawn uniformly from a bag. Counts: pink=6, red=8, cyan=2, green=4, brown=3, purple=3, gray=4, white=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"brown": "The sampled ball is brown", "green": "The sampled ball is green", "gray": "The sampled ball is gray", "cyan": "The sampled ball is cyan", "pink": "The sampled ball is pink", "purple": "The sampled ball is purple", "red": "The sampled ball is red", "white": "The sampled ball is white"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: pink, green, gray", "The drawn color is one of: red, brown, white", "The drawn color is one of: cyan, purple"]}}, "gold": {}, "gold_probs": {"outcome": {"brown": 0.08571428571428572, "green": 0.11428571428571428, "gray": 0.11428571428571428, "cyan": 0.05714285714285714, "pink": 0.17142857142857143, "purple": 0.08571428571428572, "red": 0.22857142857142856, "white": 0.14285714285714285}, "focus": {"false": 0.9428571428571428, "true": 0.05714285714285714}, "group": {"0": 0.4, "1": 0.45714285714285713, "2": 0.14285714285714285}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"3/35\"],[\"cyan\",\"2/35\"],[\"gray\",\"4/35\"],[\"green\",\"4/35\"],[\"pink\",\"6/35\"],[\"purple\",\"3/35\"],[\"red\",\"8/35\"],[\"white\",\"1/7\"]]", "template_id": "known_chance_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0, "cyan": 0, "red": 0.91, "white": 0.01, "gray": 0, "pink": 0.08, "green": 0, "brown": 0}, "focus": {"false": 0.92, "true": 0.08}, "group": {"0": 0.23, "1": 0.75, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:calibration:10", "state_id": "known_chance_v2:calibration:10", "family_id": "known_chance_v2", "split": "calibration", "state": "Probability review: exactly one ball is drawn uniformly from a bag. Counts: gray=3, cyan=7, black=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "gray": "The sampled ball is gray", "black": "The sampled ball is black"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a gray ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: gray", "The drawn color is one of: cyan", "The drawn color is one of: black"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.4117647058823529, "gray": 0.17647058823529413, "black": 0.4117647058823529}, "focus": {"false": 0.8235294117647058, "true": 0.17647058823529413}, "group": {"0": 0.17647058823529413, "1": 0.4117647058823529, "2": 0.4117647058823529}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"7/17\"],[\"cyan\",\"7/17\"],[\"gray\",\"3/17\"]]", "template_id": "known_chance_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0.53, "black": 0.05, "gray": 0.42}, "focus": {"false": 0.81, "true": 0.19}, "group": {"0": 0.05, "1": 0.83, "2": 0.12}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:calibration:11", "state_id": "known_chance_v2:calibration:11", "family_id": "known_chance_v2", "split": "calibration", "state": "Probability review: exactly one ball is drawn uniformly from a bag. Counts: white=7, brown=7, cyan=3, orange=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "brown": "The sampled ball is brown", "orange": "The sampled ball is orange", "white": "The sampled ball is white"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white, orange", "The drawn color is one of: brown", "The drawn color is one of: cyan"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.12, "brown": 0.28, "orange": 0.32, "white": 0.28}, "focus": {"false": 0.88, "true": 0.12}, "group": {"0": 0.6000000000000001, "1": 0.28, "2": 0.12}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"7/25\"],[\"cyan\",\"3/25\"],[\"orange\",\"8/25\"],[\"white\",\"7/25\"]]", "template_id": "known_chance_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0.34, "orange": 0.59, "brown": 0.04, "cyan": 0.03}, "focus": {"false": 0.87, "true": 0.13}, "group": {"0": 0.98, "1": 0.01, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:calibration:12", "state_id": "known_chance_v2:calibration:12", "family_id": "known_chance_v2", "split": "calibration", "state": "Probability review: exactly one ball is drawn uniformly from a bag. Counts: blue=6, black=7, gray=2, purple=8, red=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "red": "The sampled ball is red", "gray": "The sampled ball is gray", "black": "The sampled ball is black", "purple": "The sampled ball is purple"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a purple ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue, purple", "The drawn color is one of: black, red", "The drawn color is one of: gray"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.1935483870967742, "red": 0.25806451612903225, "gray": 0.06451612903225806, "black": 0.22580645161290322, "purple": 0.25806451612903225}, "focus": {"false": 0.7419354838709677, "true": 0.25806451612903225}, "group": {"0": 0.45161290322580644, "1": 0.4838709677419355, "2": 0.06451612903225806}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"7/31\"],[\"blue\",\"6/31\"],[\"gray\",\"2/31\"],[\"purple\",\"8/31\"],[\"red\",\"8/31\"]]", "template_id": "known_chance_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"red": 0.12, "blue": 0.33, "purple": 0.26, "gray": 0.07, "black": 0.22}, "focus": {"false": 0.78, "true": 0.22}, "group": {"0": 0.34, "1": 0.63, "2": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:calibration:13", "state_id": "known_chance_v2:calibration:13", "family_id": "known_chance_v2", "split": "calibration", "state": "Probability review: exactly one ball is drawn uniformly from a bag. Counts: black=3, white=1, pink=8, red=8, blue=1, gray=3, brown=2, orange=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"orange": "The sampled ball is orange", "blue": "The sampled ball is blue", "red": "The sampled ball is red", "gray": "The sampled ball is gray", "black": "The sampled ball is black", "pink": "The sampled ball is pink", "white": "The sampled ball is white", "brown": "The sampled ball is brown"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: black, red, brown", "The drawn color is one of: white, blue, orange", "The drawn color is one of: pink, gray"]}}, "gold": {}, "gold_probs": {"outcome": {"orange": 0.23529411764705882, "blue": 0.029411764705882353, "red": 0.23529411764705882, "gray": 0.08823529411764706, "black": 0.08823529411764706, "pink": 0.23529411764705882, "white": 0.029411764705882353, "brown": 0.058823529411764705}, "focus": {"false": 0.9705882352941176, "true": 0.029411764705882353}, "group": {"0": 0.38235294117647056, "1": 0.29411764705882354, "2": 0.3235294117647059}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"3/34\"],[\"blue\",\"1/34\"],[\"brown\",\"1/17\"],[\"gray\",\"3/34\"],[\"orange\",\"4/17\"],[\"pink\",\"4/17\"],[\"red\",\"4/17\"],[\"white\",\"1/34\"]]", "template_id": "known_chance_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"brown": 0.01, "blue": 0.01, "white": 0.01, "gray": 0.01, "red": 0.1, "orange": 0.08, "pink": 0.59, "black": 0.19}, "focus": {"false": 0.94, "true": 0.06}, "group": {"0": 0.4, "1": 0.33, "2": 0.27}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:calibration:14", "state_id": "known_chance_v2:calibration:14", "family_id": "known_chance_v2", "split": "calibration", "state": "Probability review: exactly one ball is drawn uniformly from a bag. Counts: green=6, red=2, black=2, yellow=4, orange=5, blue=8, pink=9, gray=4, brown=9, cyan=8, white=5, purple=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"gray": "The sampled ball is gray", "purple": "The sampled ball is purple", "pink": "The sampled ball is pink", "orange": "The sampled ball is orange", "yellow": "The sampled ball is yellow", "cyan": "The sampled ball is cyan", "brown": "The sampled ball is brown", "white": "The sampled ball is white", "green": "The sampled ball is green", "black": "The sampled ball is black", "red": "The sampled ball is red", "blue": "The sampled ball is blue"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a orange ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: green, yellow, pink, cyan", "The drawn color is one of: red, orange, gray, white", "The drawn color is one of: black, blue, brown, purple"]}}, "gold": {}, "gold_probs": {"outcome": {"gray": 0.06153846153846154, "purple": 0.046153846153846156, "pink": 0.13846153846153847, "orange": 0.07692307692307693, "yellow": 0.06153846153846154, "cyan": 0.12307692307692308, "brown": 0.13846153846153847, "white": 0.07692307692307693, "green": 0.09230769230769231, "black": 0.03076923076923077, "red": 0.03076923076923077, "blue": 0.12307692307692308}, "focus": {"false": 0.9230769230769231, "true": 0.07692307692307693}, "group": {"0": 0.4153846153846154, "1": 0.24615384615384617, "2": 0.3384615384615385}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"2/65\"],[\"blue\",\"8/65\"],[\"brown\",\"9/65\"],[\"cyan\",\"8/65\"],[\"gray\",\"4/65\"],[\"green\",\"6/65\"],[\"orange\",\"1/13\"],[\"pink\",\"9/65\"],[\"purple\",\"3/65\"],[\"red\",\"2/65\"],[\"white\",\"1/13\"],[\"yellow\",\"4/65\"]]", "template_id": "known_chance_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.26, "gray": 0.01, "blue": 0.07, "cyan": 0.01, "green": 0.44, "white": 0.01, "yellow": 0.01, "orange": 0.01, "purple": 0.01, "red": 0.03, "brown": 0.13, "black": 0.01}, "focus": {"false": 0.86, "true": 0.14}, "group": {"0": 0.62, "1": 0.04, "2": 0.34}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:calibration:16", "state_id": "known_chance_v2:calibration:16", "family_id": "known_chance_v2", "split": "calibration", "state": "Probability review: exactly one ball is drawn uniformly from a bag. Counts: green=5, cyan=9, white=6, red=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"white": "The sampled ball is white", "red": "The sampled ball is red", "cyan": "The sampled ball is cyan", "green": "The sampled ball is green"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: green, red", "The drawn color is one of: cyan", "The drawn color is one of: white"]}}, "gold": {}, "gold_probs": {"outcome": {"white": 0.2857142857142857, "red": 0.047619047619047616, "cyan": 0.42857142857142855, "green": 0.23809523809523808}, "focus": {"false": 0.5714285714285714, "true": 0.42857142857142855}, "group": {"0": 0.2857142857142857, "1": 0.42857142857142855, "2": 0.2857142857142857}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"3/7\"],[\"green\",\"5/21\"],[\"red\",\"1/21\"],[\"white\",\"2/7\"]]", "template_id": "known_chance_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0.01, "cyan": 0.92, "green": 0.05, "red": 0.02}, "focus": {"false": 0.71, "true": 0.29}, "group": {"0": 0.02, "1": 0.92, "2": 0.06}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:calibration:15", "state_id": "known_chance_v2:calibration:15", "family_id": "known_chance_v2", "split": "calibration", "state": "Probability review: exactly one ball is drawn uniformly from a bag. Counts: cyan=2, green=6, white=9. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "white": "The sampled ball is white", "green": "The sampled ball is green"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a white ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: cyan", "The drawn color is one of: green", "The drawn color is one of: white"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.11764705882352941, "white": 0.5294117647058824, "green": 0.35294117647058826}, "focus": {"false": 0.47058823529411764, "true": 0.5294117647058824}, "group": {"0": 0.11764705882352941, "1": 0.35294117647058826, "2": 0.5294117647058824}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"2/17\"],[\"green\",\"6/17\"],[\"white\",\"9/17\"]]", "template_id": "known_chance_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0.12, "white": 0.84, "green": 0.04}, "focus": {"false": 0.64, "true": 0.36}, "group": {"0": 0.01, "1": 0.02, "2": 0.97}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:calibration:18", "state_id": "known_chance_v2:calibration:18", "family_id": "known_chance_v2", "split": "calibration", "state": "Probability review: exactly one ball is drawn uniformly from a bag. Counts: pink=4, gray=1, orange=9, white=1, red=2, black=8, brown=4, cyan=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"brown": "The sampled ball is brown", "gray": "The sampled ball is gray", "black": "The sampled ball is black", "red": "The sampled ball is red", "white": "The sampled ball is white", "pink": "The sampled ball is pink", "orange": "The sampled ball is orange", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a orange ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: pink, white, brown", "The drawn color is one of: gray, red, cyan", "The drawn color is one of: orange, black"]}}, "gold": {}, "gold_probs": {"outcome": {"brown": 0.125, "gray": 0.03125, "black": 0.25, "red": 0.0625, "white": 0.03125, "pink": 0.125, "orange": 0.28125, "cyan": 0.09375}, "focus": {"false": 0.71875, "true": 0.28125}, "group": {"0": 0.28125, "1": 0.1875, "2": 0.53125}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/4\"],[\"brown\",\"1/8\"],[\"cyan\",\"3/32\"],[\"gray\",\"1/32\"],[\"orange\",\"9/32\"],[\"pink\",\"1/8\"],[\"red\",\"1/16\"],[\"white\",\"1/32\"]]", "template_id": "known_chance_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"black": 0.04, "red": 0, "gray": 0, "orange": 0.93, "white": 0, "pink": 0.01, "cyan": 0.01, "brown": 0}, "focus": {"false": 0.75, "true": 0.25}, "group": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:calibration:19", "state_id": "known_chance_v2:calibration:19", "family_id": "known_chance_v2", "split": "calibration", "state": "Probability review: exactly one ball is drawn uniformly from a bag. Counts: black=3, brown=9, purple=7, cyan=9, orange=4, gray=9, white=8, blue=6, yellow=8, green=6, red=8, pink=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "purple": "The sampled ball is purple", "orange": "The sampled ball is orange", "gray": "The sampled ball is gray", "blue": "The sampled ball is blue", "black": "The sampled ball is black", "brown": "The sampled ball is brown", "pink": "The sampled ball is pink", "white": "The sampled ball is white", "red": "The sampled ball is red", "cyan": "The sampled ball is cyan", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: black, cyan, white, green", "The drawn color is one of: brown, orange, blue, red", "The drawn color is one of: purple, gray, yellow, pink"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.0759493670886076, "purple": 0.08860759493670886, "orange": 0.05063291139240506, "gray": 0.11392405063291139, "blue": 0.0759493670886076, "black": 0.0379746835443038, "brown": 0.11392405063291139, "pink": 0.02531645569620253, "white": 0.10126582278481013, "red": 0.10126582278481013, "cyan": 0.11392405063291139, "yellow": 0.10126582278481013}, "focus": {"false": 0.8987341772151899, "true": 0.10126582278481013}, "group": {"0": 0.3291139240506329, "1": 0.34177215189873417, "2": 0.3291139240506329}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"3/79\"],[\"blue\",\"6/79\"],[\"brown\",\"9/79\"],[\"cyan\",\"9/79\"],[\"gray\",\"9/79\"],[\"green\",\"6/79\"],[\"orange\",\"4/79\"],[\"pink\",\"2/79\"],[\"purple\",\"7/79\"],[\"red\",\"8/79\"],[\"white\",\"8/79\"],[\"yellow\",\"8/79\"]]", "template_id": "known_chance_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.01, "cyan": 0.11, "white": 0.01, "pink": 0.09, "yellow": 0.01, "blue": 0.01, "brown": 0.46, "gray": 0.08, "red": 0.02, "orange": 0.01, "green": 0.01, "black": 0.18}, "focus": {"false": 0.86, "true": 0.14}, "group": {"0": 0.16, "1": 0.52, "2": 0.32}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:calibration:17", "state_id": "known_chance_v2:calibration:17", "family_id": "known_chance_v2", "split": "calibration", "state": "Probability review: exactly one ball is drawn uniformly from a bag. Counts: pink=9, green=3, white=9, purple=7, brown=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "white": "The sampled ball is white", "purple": "The sampled ball is purple", "brown": "The sampled ball is brown", "pink": "The sampled ball is pink"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a pink ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: pink, purple", "The drawn color is one of: green, brown", "The drawn color is one of: white"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.10344827586206896, "white": 0.3103448275862069, "purple": 0.2413793103448276, "brown": 0.034482758620689655, "pink": 0.3103448275862069}, "focus": {"false": 0.6896551724137931, "true": 0.3103448275862069}, "group": {"0": 0.5517241379310345, "1": 0.13793103448275862, "2": 0.3103448275862069}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"1/29\"],[\"green\",\"3/29\"],[\"pink\",\"9/29\"],[\"purple\",\"7/29\"],[\"white\",\"9/29\"]]", "template_id": "known_chance_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0.02, "pink": 0.76, "white": 0.14, "brown": 0.04, "purple": 0.04}, "focus": {"false": 0.76, "true": 0.24}, "group": {"0": 0.96, "1": 0.01, "2": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:calibration:21", "state_id": "known_chance_v2:calibration:21", "family_id": "known_chance_v2", "split": "calibration", "state": "Probability review: exactly one ball is drawn uniformly from a bag. Counts: black=6, blue=7, yellow=7, white=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"yellow": "The sampled ball is yellow", "black": "The sampled ball is black", "blue": "The sampled ball is blue", "white": "The sampled ball is white"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: black, white", "The drawn color is one of: blue", "The drawn color is one of: yellow"]}}, "gold": {}, "gold_probs": {"outcome": {"yellow": 0.3333333333333333, "black": 0.2857142857142857, "blue": 0.3333333333333333, "white": 0.047619047619047616}, "focus": {"false": 0.6666666666666667, "true": 0.3333333333333333}, "group": {"0": 0.3333333333333333, "1": 0.3333333333333333, "2": 0.3333333333333333}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"2/7\"],[\"blue\",\"1/3\"],[\"white\",\"1/21\"],[\"yellow\",\"1/3\"]]", "template_id": "known_chance_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"yellow": 0.06, "white": 0.05, "blue": 0.23, "black": 0.66}, "focus": {"false": 0.8, "true": 0.2}, "group": {"0": 0.27, "1": 0.55, "2": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:calibration:22", "state_id": "known_chance_v2:calibration:22", "family_id": "known_chance_v2", "split": "calibration", "state": "Probability review: exactly one ball is drawn uniformly from a bag. Counts: red=6, black=8, yellow=1, white=4, purple=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"white": "The sampled ball is white", "black": "The sampled ball is black", "yellow": "The sampled ball is yellow", "purple": "The sampled ball is purple", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a white ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: red, white", "The drawn color is one of: black, purple", "The drawn color is one of: yellow"]}}, "gold": {}, "gold_probs": {"outcome": {"white": 0.17391304347826086, "black": 0.34782608695652173, "yellow": 0.043478260869565216, "purple": 0.17391304347826086, "red": 0.2608695652173913}, "focus": {"false": 0.8260869565217391, "true": 0.17391304347826086}, "group": {"0": 0.43478260869565216, "1": 0.5217391304347826, "2": 0.043478260869565216}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"8/23\"],[\"purple\",\"4/23\"],[\"red\",\"6/23\"],[\"white\",\"4/23\"],[\"yellow\",\"1/23\"]]", "template_id": "known_chance_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"yellow": 0.03, "white": 0.01, "red": 0.1, "purple": 0, "black": 0.86}, "focus": {"false": 0.81, "true": 0.19}, "group": {"0": 0.07, "1": 0.92, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:calibration:23", "state_id": "known_chance_v2:calibration:23", "family_id": "known_chance_v2", "split": "calibration", "state": "Probability review: exactly one ball is drawn uniformly from a bag. Counts: orange=5, purple=9, yellow=9, blue=5, red=9, green=9, pink=4, brown=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"red": "The sampled ball is red", "yellow": "The sampled ball is yellow", "blue": "The sampled ball is blue", "pink": "The sampled ball is pink", "green": "The sampled ball is green", "orange": "The sampled ball is orange", "brown": "The sampled ball is brown", "purple": "The sampled ball is purple"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a red ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: orange, blue, pink", "The drawn color is one of: purple, red, brown", "The drawn color is one of: yellow, green"]}}, "gold": {}, "gold_probs": {"outcome": {"red": 0.16071428571428573, "yellow": 0.16071428571428573, "blue": 0.08928571428571429, "pink": 0.07142857142857142, "green": 0.16071428571428573, "orange": 0.08928571428571429, "brown": 0.10714285714285714, "purple": 0.16071428571428573}, "focus": {"false": 0.8392857142857143, "true": 0.16071428571428573}, "group": {"0": 0.25, "1": 0.4285714285714286, "2": 0.32142857142857145}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"5/56\"],[\"brown\",\"3/28\"],[\"green\",\"9/56\"],[\"orange\",\"5/56\"],[\"pink\",\"1/14\"],[\"purple\",\"9/56\"],[\"red\",\"9/56\"],[\"yellow\",\"9/56\"]]", "template_id": "known_chance_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.06, "yellow": 0.04, "green": 0.03, "brown": 0.02, "orange": 0.17, "purple": 0.61, "blue": 0.01, "red": 0.06}, "focus": {"false": 0.8, "true": 0.2}, "group": {"0": 0.05, "1": 0.71, "2": 0.24}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:calibration:24", "state_id": "known_chance_v2:calibration:24", "family_id": "known_chance_v2", "split": "calibration", "state": "Probability review: exactly one ball is drawn uniformly from a bag. Counts: green=1, black=7, yellow=9, orange=1, purple=7, red=6, blue=9, gray=6, pink=7, white=1, brown=8, cyan=9. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"white": "The sampled ball is white", "yellow": "The sampled ball is yellow", "gray": "The sampled ball is gray", "cyan": "The sampled ball is cyan", "purple": "The sampled ball is purple", "brown": "The sampled ball is brown", "green": "The sampled ball is green", "blue": "The sampled ball is blue", "orange": "The sampled ball is orange", "pink": "The sampled ball is pink", "black": "The sampled ball is black", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a green ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: green, orange, blue, white", "The drawn color is one of: black, purple, gray, brown", "The drawn color is one of: yellow, red, pink, cyan"]}}, "gold": {}, "gold_probs": {"outcome": {"white": 0.014084507042253521, "yellow": 0.1267605633802817, "gray": 0.08450704225352113, "cyan": 0.1267605633802817, "purple": 0.09859154929577464, "brown": 0.11267605633802817, "green": 0.014084507042253521, "blue": 0.1267605633802817, "orange": 0.014084507042253521, "pink": 0.09859154929577464, "black": 0.09859154929577464, "red": 0.08450704225352113}, "focus": {"false": 0.9859154929577465, "true": 0.014084507042253521}, "group": {"0": 0.16901408450704225, "1": 0.39436619718309857, "2": 0.43661971830985913}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"7/71\"],[\"blue\",\"9/71\"],[\"brown\",\"8/71\"],[\"cyan\",\"9/71\"],[\"gray\",\"6/71\"],[\"green\",\"1/71\"],[\"orange\",\"1/71\"],[\"pink\",\"7/71\"],[\"purple\",\"7/71\"],[\"red\",\"6/71\"],[\"white\",\"1/71\"],[\"yellow\",\"9/71\"]]", "template_id": "known_chance_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0.07, "orange": 0.01, "red": 0.01, "purple": 0.01, "pink": 0.01, "gray": 0.01, "brown": 0.03, "yellow": 0.44, "cyan": 0.09, "white": 0.01, "blue": 0.21, "black": 0.1}, "focus": {"false": 0.94, "true": 0.06}, "group": {"0": 0.08, "1": 0.22, "2": 0.7}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:calibration:20", "state_id": "known_chance_v2:calibration:20", "family_id": "known_chance_v2", "split": "calibration", "state": "Probability review: exactly one ball is drawn uniformly from a bag. Counts: green=8, red=1, pink=9. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "red": "The sampled ball is red", "pink": "The sampled ball is pink"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a pink ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: green", "The drawn color is one of: red", "The drawn color is one of: pink"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.4444444444444444, "red": 0.05555555555555555, "pink": 0.5}, "focus": {"false": 0.5, "true": 0.5}, "group": {"0": 0.4444444444444444, "1": 0.05555555555555555, "2": 0.5}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"green\",\"4/9\"],[\"pink\",\"1/2\"],[\"red\",\"1/18\"]]", "template_id": "known_chance_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"red": 0.03, "green": 0.36, "pink": 0.61}, "focus": {"false": 0.63, "true": 0.37}, "group": {"0": 0.08, "1": 0.01, "2": 0.91}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:calibration:27", "state_id": "known_chance_v2:calibration:27", "family_id": "known_chance_v2", "split": "calibration", "state": "Probability review: exactly one ball is drawn uniformly from a bag. Counts: black=5, orange=3, green=6, brown=9, blue=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "orange": "The sampled ball is orange", "blue": "The sampled ball is blue", "brown": "The sampled ball is brown", "green": "The sampled ball is green"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: black, brown", "The drawn color is one of: orange, blue", "The drawn color is one of: green"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.17857142857142858, "orange": 0.10714285714285714, "blue": 0.17857142857142858, "brown": 0.32142857142857145, "green": 0.21428571428571427}, "focus": {"false": 0.8214285714285714, "true": 0.17857142857142858}, "group": {"0": 0.5, "1": 0.2857142857142857, "2": 0.21428571428571427}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"5/28\"],[\"blue\",\"5/28\"],[\"brown\",\"9/28\"],[\"green\",\"3/14\"],[\"orange\",\"3/28\"]]", "template_id": "known_chance_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"black": 0.06, "brown": 0.91, "blue": 0.01, "green": 0.01, "orange": 0.01}, "focus": {"false": 0.8200000000000001, "true": 0.18}, "group": {"0": 0.96, "1": 0.01, "2": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:calibration:26", "state_id": "known_chance_v2:calibration:26", "family_id": "known_chance_v2", "split": "calibration", "state": "Probability review: exactly one ball is drawn uniformly from a bag. Counts: pink=8, cyan=9, orange=5, green=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"pink": "The sampled ball is pink", "cyan": "The sampled ball is cyan", "green": "The sampled ball is green", "orange": "The sampled ball is orange"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a pink ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: pink, green", "The drawn color is one of: cyan", "The drawn color is one of: orange"]}}, "gold": {}, "gold_probs": {"outcome": {"pink": 0.26666666666666666, "cyan": 0.3, "green": 0.26666666666666666, "orange": 0.16666666666666666}, "focus": {"false": 0.7333333333333334, "true": 0.26666666666666666}, "group": {"0": 0.5333333333333333, "1": 0.3, "2": 0.16666666666666666}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"3/10\"],[\"green\",\"4/15\"],[\"orange\",\"1/6\"],[\"pink\",\"4/15\"]]", "template_id": "known_chance_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"orange": 0.01, "green": 0.01, "pink": 0.17, "cyan": 0.81}, "focus": {"false": 0.77, "true": 0.23}, "group": {"0": 0.33, "1": 0.66, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:calibration:29", "state_id": "known_chance_v2:calibration:29", "family_id": "known_chance_v2", "split": "calibration", "state": "Probability review: exactly one ball is drawn uniformly from a bag. Counts: cyan=8, white=5, brown=7, pink=3, gray=6, blue=2, orange=8, black=2, yellow=6, red=3, purple=8, green=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"orange": "The sampled ball is orange", "blue": "The sampled ball is blue", "black": "The sampled ball is black", "red": "The sampled ball is red", "gray": "The sampled ball is gray", "pink": "The sampled ball is pink", "yellow": "The sampled ball is yellow", "cyan": "The sampled ball is cyan", "white": "The sampled ball is white", "purple": "The sampled ball is purple", "brown": "The sampled ball is brown", "green": "The sampled ball is green"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: cyan, pink, orange, red", "The drawn color is one of: white, gray, black, purple", "The drawn color is one of: brown, blue, yellow, green"]}}, "gold": {}, "gold_probs": {"outcome": {"orange": 0.12903225806451613, "blue": 0.03225806451612903, "black": 0.03225806451612903, "red": 0.04838709677419355, "gray": 0.0967741935483871, "pink": 0.04838709677419355, "yellow": 0.0967741935483871, "cyan": 0.12903225806451613, "white": 0.08064516129032258, "purple": 0.12903225806451613, "brown": 0.11290322580645161, "green": 0.06451612903225806}, "focus": {"false": 0.967741935483871, "true": 0.03225806451612903}, "group": {"0": 0.3548387096774194, "1": 0.3387096774193548, "2": 0.3064516129032258}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/31\"],[\"blue\",\"1/31\"],[\"brown\",\"7/62\"],[\"cyan\",\"4/31\"],[\"gray\",\"3/31\"],[\"green\",\"2/31\"],[\"orange\",\"4/31\"],[\"pink\",\"3/62\"],[\"purple\",\"4/31\"],[\"red\",\"3/62\"],[\"white\",\"5/62\"],[\"yellow\",\"3/31\"]]", "template_id": "known_chance_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"blue": 0.01, "red": 0.01, "white": 0.01, "orange": 0.07, "green": 0.01, "black": 0.01, "gray": 0.02, "yellow": 0.01, "brown": 0.03, "purple": 0.06, "pink": 0.01, "cyan": 0.75}, "focus": {"false": 0.9, "true": 0.1}, "group": {"0": 0.55, "1": 0.21, "2": 0.24}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:calibration:30", "state_id": "known_chance_v2:calibration:30", "family_id": "known_chance_v2", "split": "calibration", "state": "Probability review: exactly one ball is drawn uniformly from a bag. Counts: red=1, gray=4, white=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"white": "The sampled ball is white", "red": "The sampled ball is red", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a gray ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: red", "The drawn color is one of: gray", "The drawn color is one of: white"]}}, "gold": {}, "gold_probs": {"outcome": {"white": 0.5833333333333334, "red": 0.08333333333333333, "gray": 0.3333333333333333}, "focus": {"false": 0.6666666666666667, "true": 0.3333333333333333}, "group": {"0": 0.08333333333333333, "1": 0.3333333333333333, "2": 0.5833333333333334}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"gray\",\"1/3\"],[\"red\",\"1/12\"],[\"white\",\"7/12\"]]", "template_id": "known_chance_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0.79, "gray": 0.04, "red": 0.17}, "focus": {"false": 0.72, "true": 0.28}, "group": {"0": 0.02, "1": 0.03, "2": 0.95}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:calibration:28", "state_id": "known_chance_v2:calibration:28", "family_id": "known_chance_v2", "split": "calibration", "state": "Probability review: exactly one ball is drawn uniformly from a bag. Counts: orange=9, red=4, gray=9, pink=6, cyan=8, blue=9, white=8, brown=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"gray": "The sampled ball is gray", "red": "The sampled ball is red", "pink": "The sampled ball is pink", "cyan": "The sampled ball is cyan", "orange": "The sampled ball is orange", "white": "The sampled ball is white", "brown": "The sampled ball is brown", "blue": "The sampled ball is blue"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a white ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: orange, pink, white", "The drawn color is one of: red, cyan, brown", "The drawn color is one of: gray, blue"]}}, "gold": {}, "gold_probs": {"outcome": {"gray": 0.16071428571428573, "red": 0.07142857142857142, "pink": 0.10714285714285714, "cyan": 0.14285714285714285, "orange": 0.16071428571428573, "white": 0.14285714285714285, "brown": 0.05357142857142857, "blue": 0.16071428571428573}, "focus": {"false": 0.8571428571428572, "true": 0.14285714285714285}, "group": {"0": 0.4107142857142857, "1": 0.26785714285714285, "2": 0.32142857142857145}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"9/56\"],[\"brown\",\"3/56\"],[\"cyan\",\"1/7\"],[\"gray\",\"9/56\"],[\"orange\",\"9/56\"],[\"pink\",\"3/28\"],[\"red\",\"1/14\"],[\"white\",\"1/7\"]]", "template_id": "known_chance_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.01, "orange": 0.8, "blue": 0.05, "red": 0.02, "cyan": 0.01, "white": 0.01, "brown": 0.06, "gray": 0.04}, "focus": {"false": 0.86, "true": 0.14}, "group": {"0": 0.31, "1": 0.04, "2": 0.65}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:calibration:25", "state_id": "known_chance_v2:calibration:25", "family_id": "known_chance_v2", "split": "calibration", "state": "Probability review: exactly one ball is drawn uniformly from a bag. Counts: purple=7, black=4, gray=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"purple": "The sampled ball is purple", "black": "The sampled ball is black", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a gray ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: purple", "The drawn color is one of: black", "The drawn color is one of: gray"]}}, "gold": {}, "gold_probs": {"outcome": {"purple": 0.5, "black": 0.2857142857142857, "gray": 0.21428571428571427}, "focus": {"false": 0.7857142857142857, "true": 0.21428571428571427}, "group": {"0": 0.5, "1": 0.2857142857142857, "2": 0.21428571428571427}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"2/7\"],[\"gray\",\"3/14\"],[\"purple\",\"1/2\"]]", "template_id": "known_chance_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"black": 0.01, "purple": 0.98, "gray": 0.01}, "focus": {"false": 0.81, "true": 0.19}, "group": {"0": 0.97, "1": 0.01, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:0", "state_id": "smart_home_v2:test:0", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the dining room fan set to on; access=yes; occupants=2; clock=20:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony speaker", "unit_1": "dining room fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"dining room\",\"fan\",true,2]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 0.040000000000000036, "true": 0.96}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:2", "state_id": "smart_home_v2:test:2", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the office fan set to off; access=yes; occupants=3; clock=17:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom speaker", "unit_1": "garage fan", "unit_2": "balcony fan", "unit_3": "office fan", "unit_4": "library speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_3", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 1, "unit_4": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"office\",\"fan\",true,3]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_4": 0, "unit_3": 1}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:calibration:31", "state_id": "known_chance_v2:calibration:31", "family_id": "known_chance_v2", "split": "calibration", "state": "Probability review: exactly one ball is drawn uniformly from a bag. Counts: purple=8, gray=2, yellow=6, red=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"purple": "The sampled ball is purple", "red": "The sampled ball is red", "gray": "The sampled ball is gray", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: purple, red", "The drawn color is one of: gray", "The drawn color is one of: yellow"]}}, "gold": {}, "gold_probs": {"outcome": {"purple": 0.38095238095238093, "red": 0.23809523809523808, "gray": 0.09523809523809523, "yellow": 0.2857142857142857}, "focus": {"false": 0.7142857142857143, "true": 0.2857142857142857}, "group": {"0": 0.6190476190476191, "1": 0.09523809523809523, "2": 0.2857142857142857}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"gray\",\"2/21\"],[\"purple\",\"8/21\"],[\"red\",\"5/21\"],[\"yellow\",\"2/7\"]]", "template_id": "known_chance_v2:calibration"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"yellow": 0.01, "red": 0.01, "purple": 0.97, "gray": 0.01}, "focus": {"false": 0.79, "true": 0.21}, "group": {"0": 0.97, "1": 0.01, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:1", "state_id": "smart_home_v2:test:1", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the bedroom heater set to off; access=yes; occupants=0; clock=8:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bathroom lamp", "unit_1": "bedroom heater", "unit_2": "garden heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 1}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 1, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bedroom\",\"heater\",true,0]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0}, "execute": {"false": 0.95, "true": 0.05}, "risk": {"0": 0, "1": 1, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:3", "state_id": "smart_home_v2:test:3", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the library lamp set to off; access=yes; occupants=3; clock=3:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony heater", "unit_1": "kitchen lamp", "unit_2": "dining room speaker", "unit_3": "bathroom heater", "unit_4": "garage speaker", "unit_5": "library speaker", "unit_6": "kitchen speaker", "unit_7": "library lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_7", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"library\",\"lamp\",true,3]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_3": 0, "unit_4": 0, "unit_2": 0, "unit_0": 0, "unit_6": 0, "unit_7": 1, "unit_5": 0, "unit_1": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:4", "state_id": "smart_home_v2:test:4", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the balcony speaker set to on; access=no; occupants=2; clock=20:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "attic speaker", "unit_1": "office fan", "unit_2": "office heater", "unit_3": "office lamp", "unit_4": "hall lamp", "unit_5": "balcony speaker", "unit_6": "kitchen speaker", "unit_7": "dining room heater", "unit_8": "balcony fan", "unit_9": "bedroom fan", "unit_10": "hall heater", "unit_11": "dining room speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_5", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 1, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"balcony\",\"speaker\",false,2]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_4": 0, "unit_0": 0, "unit_9": 0, "unit_3": 0, "unit_7": 0, "unit_2": 0, "unit_10": 0, "unit_6": 0, "unit_5": 1, "unit_1": 0, "unit_11": 0, "unit_8": 0}, "execute": {"false": 0.95, "true": 0.05}, "risk": {"0": 0.02, "1": 0, "2": 0.98}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:6", "state_id": "smart_home_v2:test:6", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the attic speaker set to on; access=yes; occupants=3; clock=10:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "attic speaker", "unit_1": "dining room fan", "unit_2": "garden speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"attic\",\"speaker\",true,3]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 0.06999999999999995, "true": 0.93}, "risk": {"0": 0.99, "1": 0, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:5", "state_id": "smart_home_v2:test:5", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the garden speaker set to on; access=no; occupants=3; clock=11:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom heater", "unit_1": "garden speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garden\",\"speaker\",false,3]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 0.95, "true": 0.05}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:9", "state_id": "smart_home_v2:test:9", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the dining room heater set to off; access=no; occupants=1; clock=9:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library speaker", "unit_1": "dining room speaker", "unit_2": "hall heater", "unit_3": "garden speaker", "unit_4": "dining room heater", "unit_5": "bedroom speaker", "unit_6": "library fan", "unit_7": "garage speaker", "unit_8": "living room heater", "unit_9": "living room speaker", "unit_10": "kitchen lamp", "unit_11": "garden lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_4", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 1, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"dining room\",\"heater\",false,1]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_3": 0, "unit_8": 0, "unit_1": 0, "unit_11": 0, "unit_7": 0, "unit_2": 0, "unit_0": 0, "unit_9": 0, "unit_10": 0, "unit_6": 0, "unit_5": 0, "unit_4": 1}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.06, "1": 0, "2": 0.94}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:11", "state_id": "smart_home_v2:test:11", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the balcony fan set to on; access=yes; occupants=2; clock=21:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "attic heater", "unit_1": "hall lamp", "unit_2": "balcony fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"balcony\",\"fan\",true,2]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 1, "unit_1": 0, "unit_0": 0}, "execute": {"false": 0.040000000000000036, "true": 0.96}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:10", "state_id": "smart_home_v2:test:10", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the garden heater set to off; access=no; occupants=1; clock=1:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garden heater", "unit_1": "bathroom fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garden\",\"heater\",false,1]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.03, "1": 0, "2": 0.97}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:7", "state_id": "smart_home_v2:test:7", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the living room lamp set to off; access=yes; occupants=1; clock=3:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room lamp", "unit_1": "bedroom speaker", "unit_2": "bathroom fan", "unit_3": "garden fan", "unit_4": "garden speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"living room\",\"lamp\",true,1]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_0": 1, "unit_3": 0, "unit_4": 0, "unit_1": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:14", "state_id": "smart_home_v2:test:14", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the balcony speaker set to off; access=no; occupants=1; clock=3:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "hall heater", "unit_1": "garden lamp", "unit_2": "hall lamp", "unit_3": "library fan", "unit_4": "attic fan", "unit_5": "balcony speaker", "unit_6": "living room lamp", "unit_7": "kitchen speaker", "unit_8": "bathroom speaker", "unit_9": "hall speaker", "unit_10": "living room speaker", "unit_11": "bathroom lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_5", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 1, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"balcony\",\"speaker\",false,1]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_4": 0, "unit_5": 1, "unit_8": 0, "unit_2": 0, "unit_9": 0, "unit_7": 0, "unit_1": 0, "unit_3": 0, "unit_6": 0, "unit_11": 0, "unit_10": 0, "unit_0": 0}, "execute": {"false": 0.96, "true": 0.04}, "risk": {"0": 0.03, "1": 0, "2": 0.97}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:12", "state_id": "smart_home_v2:test:12", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the garden heater set to off; access=yes; occupants=2; clock=15:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bathroom speaker", "unit_1": "office fan", "unit_2": "garden heater", "unit_3": "kitchen speaker", "unit_4": "living room lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1, "unit_3": 0, "unit_4": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garden\",\"heater\",true,2]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_3": 0, "unit_4": 0, "unit_0": 0, "unit_1": 0, "unit_2": 1}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:15", "state_id": "smart_home_v2:test:15", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the living room speaker set to on; access=yes; occupants=1; clock=12:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "kitchen speaker", "unit_1": "living room speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"living room\",\"speaker\",true,1]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 0.040000000000000036, "true": 0.96}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:8", "state_id": "smart_home_v2:test:8", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the hall speaker set to off; access=no; occupants=0; clock=12:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony lamp", "unit_1": "bathroom lamp", "unit_2": "balcony speaker", "unit_3": "kitchen heater", "unit_4": "bedroom speaker", "unit_5": "hall speaker", "unit_6": "living room fan", "unit_7": "garage speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_5", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 1, "unit_6": 0, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"hall\",\"speaker\",false,0]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_5": 1, "unit_0": 0, "unit_2": 0, "unit_7": 0, "unit_4": 0, "unit_3": 0, "unit_6": 0}, "execute": {"false": 0.95, "true": 0.05}, "risk": {"0": 0.02, "1": 0, "2": 0.98}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:16", "state_id": "smart_home_v2:test:16", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the bedroom speaker set to off; access=no; occupants=0; clock=18:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom speaker", "unit_1": "hall speaker", "unit_2": "office fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bedroom\",\"speaker\",false,0]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 0.95, "true": 0.05}, "risk": {"0": 0.02, "1": 0, "2": 0.98}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:13", "state_id": "smart_home_v2:test:13", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the library fan set to on; access=yes; occupants=3; clock=3:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "hall lamp", "unit_1": "office fan", "unit_2": "attic speaker", "unit_3": "office lamp", "unit_4": "dining room lamp", "unit_5": "library fan", "unit_6": "balcony lamp", "unit_7": "bathroom speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_5", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 1, "unit_6": 0, "unit_7": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"library\",\"fan\",true,3]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_7": 0, "unit_2": 0, "unit_3": 0, "unit_5": 1, "unit_6": 0, "unit_4": 0, "unit_1": 0, "unit_0": 0}, "execute": {"false": 0.06999999999999995, "true": 0.93}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:17", "state_id": "smart_home_v2:test:17", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the office heater set to off; access=yes; occupants=3; clock=6:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garden fan", "unit_1": "office heater", "unit_2": "dining room speaker", "unit_3": "garage speaker", "unit_4": "garden speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0, "unit_3": 0, "unit_4": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"office\",\"heater\",true,3]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_3": 0, "unit_1": 1, "unit_4": 0, "unit_2": 0, "unit_0": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:19", "state_id": "smart_home_v2:test:19", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the hall lamp set to on; access=no; occupants=3; clock=13:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garden lamp", "unit_1": "kitchen fan", "unit_2": "balcony speaker", "unit_3": "bedroom speaker", "unit_4": "garden heater", "unit_5": "kitchen lamp", "unit_6": "bathroom fan", "unit_7": "dining room lamp", "unit_8": "hall lamp", "unit_9": "hall heater", "unit_10": "balcony heater", "unit_11": "library speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_8", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 1, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"hall\",\"lamp\",false,3]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_5": 0, "unit_10": 0, "unit_4": 0, "unit_11": 0, "unit_3": 0, "unit_2": 0, "unit_1": 0, "unit_6": 0, "unit_9": 0, "unit_7": 0, "unit_0": 0, "unit_8": 1}, "execute": {"false": 0.96, "true": 0.04}, "risk": {"0": 0.02, "1": 0, "2": 0.98}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:18", "state_id": "smart_home_v2:test:18", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the bedroom heater set to on; access=no; occupants=1; clock=0:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bathroom speaker", "unit_1": "bathroom lamp", "unit_2": "garage lamp", "unit_3": "bedroom heater", "unit_4": "attic fan", "unit_5": "living room heater", "unit_6": "hall speaker", "unit_7": "balcony speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_3", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 1, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bedroom\",\"heater\",false,1]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_3": 1, "unit_6": 0, "unit_7": 0, "unit_0": 0, "unit_4": 0, "unit_5": 0, "unit_1": 0, "unit_2": 0}, "execute": {"false": 0.96, "true": 0.04}, "risk": {"0": 0.02, "1": 0, "2": 0.98}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:20", "state_id": "smart_home_v2:test:20", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the attic lamp set to off; access=no; occupants=0; clock=5:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "office heater", "unit_1": "attic lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"attic\",\"lamp\",false,0]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 0.96, "true": 0.04}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:21", "state_id": "smart_home_v2:test:21", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the balcony heater set to on; access=yes; occupants=1; clock=13:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bathroom speaker", "unit_1": "garage heater", "unit_2": "balcony heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"balcony\",\"heater\",true,1]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_0": 0, "unit_2": 1}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:23", "state_id": "smart_home_v2:test:23", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the living room lamp set to on; access=no; occupants=1; clock=5:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room lamp", "unit_1": "library lamp", "unit_2": "living room speaker", "unit_3": "bedroom speaker", "unit_4": "library fan", "unit_5": "balcony heater", "unit_6": "kitchen lamp", "unit_7": "bathroom heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"living room\",\"lamp\",false,1]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_4": 0, "unit_7": 0, "unit_5": 0, "unit_2": 0, "unit_1": 0, "unit_3": 0, "unit_6": 0}, "execute": {"false": 0.95, "true": 0.05}, "risk": {"0": 0.02, "1": 0, "2": 0.98}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:22", "state_id": "smart_home_v2:test:22", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the kitchen speaker set to off; access=no; occupants=2; clock=3:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "attic heater", "unit_1": "kitchen speaker", "unit_2": "library lamp", "unit_3": "kitchen lamp", "unit_4": "dining room speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0, "unit_3": 0, "unit_4": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"kitchen\",\"speaker\",false,2]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_2": 0, "unit_1": 1, "unit_4": 0, "unit_3": 0}, "execute": {"false": 0.96, "true": 0.04}, "risk": {"0": 0.03, "1": 0, "2": 0.97}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:24", "state_id": "smart_home_v2:test:24", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the balcony speaker set to off; access=yes; occupants=0; clock=14:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "kitchen lamp", "unit_1": "library fan", "unit_2": "living room fan", "unit_3": "bedroom fan", "unit_4": "balcony speaker", "unit_5": "dining room lamp", "unit_6": "attic speaker", "unit_7": "dining room speaker", "unit_8": "attic heater", "unit_9": "office heater", "unit_10": "office speaker", "unit_11": "hall lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_4", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 1, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"balcony\",\"speaker\",true,0]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_7": 0, "unit_0": 0, "unit_5": 0, "unit_10": 0, "unit_8": 0, "unit_3": 0, "unit_11": 0, "unit_4": 1, "unit_2": 0, "unit_6": 0, "unit_9": 0}, "execute": {"false": 0.040000000000000036, "true": 0.96}, "risk": {"0": 0.97, "1": 0.02, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:25", "state_id": "smart_home_v2:test:25", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the living room speaker set to on; access=no; occupants=0; clock=2:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room speaker", "unit_1": "balcony speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"living room\",\"speaker\",false,0]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0.96, "true": 0.04}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:26", "state_id": "smart_home_v2:test:26", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the balcony speaker set to off; access=no; occupants=0; clock=12:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "dining room speaker", "unit_1": "garden speaker", "unit_2": "balcony speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"balcony\",\"speaker\",false,0]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_2": 1, "unit_1": 0}, "execute": {"false": 0.95, "true": 0.05}, "risk": {"0": 0.02, "1": 0, "2": 0.98}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:28", "state_id": "smart_home_v2:test:28", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the hall fan set to off; access=yes; occupants=2; clock=1:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bathroom heater", "unit_1": "garage speaker", "unit_2": "dining room lamp", "unit_3": "garden lamp", "unit_4": "hall heater", "unit_5": "garden heater", "unit_6": "garage fan", "unit_7": "hall fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_7", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"hall\",\"fan\",true,2]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_7": 1, "unit_2": 0, "unit_3": 0, "unit_5": 0, "unit_6": 0, "unit_4": 0, "unit_1": 0, "unit_0": 0}, "execute": {"false": 0.040000000000000036, "true": 0.96}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:27", "state_id": "smart_home_v2:test:27", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the kitchen lamp set to off; access=yes; occupants=0; clock=17:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room speaker", "unit_1": "kitchen lamp", "unit_2": "bedroom heater", "unit_3": "kitchen speaker", "unit_4": "office fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0, "unit_3": 0, "unit_4": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"kitchen\",\"lamp\",true,0]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_3": 0, "unit_2": 0, "unit_1": 1, "unit_4": 0, "unit_0": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:29", "state_id": "smart_home_v2:test:29", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the bathroom heater set to off; access=yes; occupants=3; clock=20:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garden heater", "unit_1": "bathroom speaker", "unit_2": "living room heater", "unit_3": "garden lamp", "unit_4": "library lamp", "unit_5": "office speaker", "unit_6": "bathroom lamp", "unit_7": "dining room speaker", "unit_8": "bathroom heater", "unit_9": "balcony fan", "unit_10": "library fan", "unit_11": "balcony heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_8", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 1, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bathroom\",\"heater\",true,3]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_2": 0, "unit_8": 1, "unit_5": 0, "unit_4": 0, "unit_0": 0, "unit_11": 0, "unit_7": 0, "unit_3": 0, "unit_9": 0, "unit_10": 0, "unit_6": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:30", "state_id": "smart_home_v2:test:30", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the garage lamp set to on; access=no; occupants=0; clock=17:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room fan", "unit_1": "garage lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garage\",\"lamp\",false,0]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 0.96, "true": 0.04}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:31", "state_id": "smart_home_v2:test:31", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the hall lamp set to off; access=yes; occupants=0; clock=23:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "hall lamp", "unit_1": "kitchen speaker", "unit_2": "living room lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"hall\",\"lamp\",true,0]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_0": 1, "unit_1": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:32", "state_id": "smart_home_v2:test:32", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the garden lamp set to off; access=yes; occupants=2; clock=6:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bathroom fan", "unit_1": "hall lamp", "unit_2": "garden lamp", "unit_3": "attic lamp", "unit_4": "balcony speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1, "unit_3": 0, "unit_4": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garden\",\"lamp\",true,2]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 1, "unit_3": 0, "unit_1": 0, "unit_4": 0, "unit_0": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:35", "state_id": "smart_home_v2:test:35", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the living room heater set to on; access=no; occupants=1; clock=2:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garage speaker", "unit_1": "living room heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"living room\",\"heater\",false,1]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 1, "unit_0": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:34", "state_id": "smart_home_v2:test:34", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the hall lamp set to off; access=no; occupants=0; clock=22:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "attic speaker", "unit_1": "garden heater", "unit_2": "library fan", "unit_3": "balcony lamp", "unit_4": "bedroom speaker", "unit_5": "office lamp", "unit_6": "garden lamp", "unit_7": "hall fan", "unit_8": "hall lamp", "unit_9": "office heater", "unit_10": "garage heater", "unit_11": "dining room speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_8", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 1, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"hall\",\"lamp\",false,0]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_4": 0, "unit_2": 0, "unit_8": 1, "unit_5": 0, "unit_10": 0, "unit_0": 0, "unit_11": 0, "unit_9": 0, "unit_1": 0, "unit_7": 0, "unit_6": 0, "unit_3": 0}, "execute": {"false": 0.95, "true": 0.05}, "risk": {"0": 0.02, "1": 0, "2": 0.98}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:36", "state_id": "smart_home_v2:test:36", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the library heater set to off; access=no; occupants=0; clock=13:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library fan", "unit_1": "library heater", "unit_2": "office speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"library\",\"heater\",false,0]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 1, "unit_0": 0, "unit_2": 0}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0, "1": 0.01, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:37", "state_id": "smart_home_v2:test:37", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the balcony speaker set to off; access=yes; occupants=2; clock=22:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony speaker", "unit_1": "garden speaker", "unit_2": "office heater", "unit_3": "living room speaker", "unit_4": "office lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"balcony\",\"speaker\",true,2]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_3": 0, "unit_1": 0, "unit_0": 1, "unit_4": 0}, "execute": {"false": 0.040000000000000036, "true": 0.96}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:38", "state_id": "smart_home_v2:test:38", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the attic heater set to off; access=no; occupants=1; clock=12:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "dining room fan", "unit_1": "library lamp", "unit_2": "living room lamp", "unit_3": "kitchen lamp", "unit_4": "garage lamp", "unit_5": "balcony heater", "unit_6": "library heater", "unit_7": "attic heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_7", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"attic\",\"heater\",false,1]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_3": 0, "unit_1": 0, "unit_2": 0, "unit_7": 1, "unit_0": 0, "unit_6": 0, "unit_5": 0, "unit_4": 0}, "execute": {"false": 0.96, "true": 0.04}, "risk": {"0": 0.07, "1": 0, "2": 0.93}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:41", "state_id": "smart_home_v2:test:41", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the dining room lamp set to on; access=yes; occupants=0; clock=12:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library heater", "unit_1": "dining room lamp", "unit_2": "living room lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"dining room\",\"lamp\",true,0]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0}, "execute": {"false": 0.040000000000000036, "true": 0.96}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:40", "state_id": "smart_home_v2:test:40", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the office lamp set to on; access=yes; occupants=1; clock=3:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "kitchen fan", "unit_1": "office lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"office\",\"lamp\",true,1]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 1, "unit_0": 0}, "execute": {"false": 0.040000000000000036, "true": 0.96}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:39", "state_id": "smart_home_v2:test:39", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the hall speaker set to on; access=no; occupants=1; clock=17:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "hall speaker", "unit_1": "garden fan", "unit_2": "bathroom fan", "unit_3": "bathroom lamp", "unit_4": "bathroom speaker", "unit_5": "garage heater", "unit_6": "kitchen lamp", "unit_7": "bedroom fan", "unit_8": "hall heater", "unit_9": "bedroom lamp", "unit_10": "attic lamp", "unit_11": "attic speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"hall\",\"speaker\",false,1]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_7": 0, "unit_2": 0, "unit_10": 0, "unit_0": 1, "unit_5": 0, "unit_4": 0, "unit_1": 0, "unit_11": 0, "unit_8": 0, "unit_3": 0, "unit_6": 0, "unit_9": 0}, "execute": {"false": 0.95, "true": 0.05}, "risk": {"0": 0.06, "1": 0, "2": 0.94}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:42", "state_id": "smart_home_v2:test:42", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the library heater set to on; access=no; occupants=3; clock=23:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bathroom speaker", "unit_1": "bedroom lamp", "unit_2": "kitchen speaker", "unit_3": "library heater", "unit_4": "garage fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_3", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 1, "unit_4": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"library\",\"heater\",false,3]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_3": 1, "unit_1": 0, "unit_0": 0, "unit_4": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:43", "state_id": "smart_home_v2:test:43", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the bedroom lamp set to on; access=no; occupants=2; clock=5:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom lamp", "unit_1": "dining room lamp", "unit_2": "office speaker", "unit_3": "bathroom fan", "unit_4": "kitchen heater", "unit_5": "office fan", "unit_6": "hall lamp", "unit_7": "garden fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bedroom\",\"lamp\",false,2]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_5": 0, "unit_4": 0, "unit_0": 1, "unit_3": 0, "unit_2": 0, "unit_7": 0, "unit_1": 0, "unit_6": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:45", "state_id": "smart_home_v2:test:45", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the attic heater set to off; access=no; occupants=3; clock=1:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room fan", "unit_1": "attic heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"attic\",\"heater\",false,3]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 1, "unit_0": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0.08, "1": 0, "2": 0.92}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:44", "state_id": "smart_home_v2:test:44", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the hall lamp set to off; access=yes; occupants=3; clock=18:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony heater", "unit_1": "garage speaker", "unit_2": "garden speaker", "unit_3": "bathroom heater", "unit_4": "library heater", "unit_5": "garage fan", "unit_6": "hall lamp", "unit_7": "hall fan", "unit_8": "hall speaker", "unit_9": "kitchen fan", "unit_10": "dining room lamp", "unit_11": "dining room speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_6", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 1, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"hall\",\"lamp\",true,3]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_4": 0, "unit_3": 0, "unit_9": 0, "unit_0": 0, "unit_7": 0, "unit_6": 1, "unit_10": 0, "unit_2": 0, "unit_8": 0, "unit_1": 0, "unit_11": 0, "unit_5": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:33", "state_id": "smart_home_v2:test:33", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the bathroom lamp set to on; access=yes; occupants=2; clock=7:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony speaker", "unit_1": "hall heater", "unit_2": "balcony heater", "unit_3": "living room heater", "unit_4": "bathroom lamp", "unit_5": "garden speaker", "unit_6": "hall fan", "unit_7": "dining room lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_4", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 1, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bathroom\",\"lamp\",true,2]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_3": 0, "unit_4": 1, "unit_2": 0, "unit_0": 0, "unit_6": 0, "unit_7": 0, "unit_5": 0, "unit_1": 0}, "execute": {"false": 0.040000000000000036, "true": 0.96}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:46", "state_id": "smart_home_v2:test:46", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the hall fan set to on; access=no; occupants=3; clock=14:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library fan", "unit_1": "hall fan", "unit_2": "library heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"hall\",\"fan\",false,3]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 1, "unit_0": 0, "unit_2": 0}, "execute": {"false": 0.96, "true": 0.04}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:48", "state_id": "smart_home_v2:test:48", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the bathroom lamp set to on; access=no; occupants=1; clock=4:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "kitchen fan", "unit_1": "balcony lamp", "unit_2": "bathroom lamp", "unit_3": "office heater", "unit_4": "living room fan", "unit_5": "dining room fan", "unit_6": "balcony speaker", "unit_7": "dining room lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bathroom\",\"lamp\",false,1]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_5": 0, "unit_7": 0, "unit_6": 0, "unit_4": 0, "unit_0": 0, "unit_1": 0, "unit_2": 1, "unit_3": 0}, "execute": {"false": 0.96, "true": 0.04}, "risk": {"0": 0.02, "1": 0, "2": 0.98}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:47", "state_id": "smart_home_v2:test:47", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the kitchen speaker set to off; access=yes; occupants=2; clock=15:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom fan", "unit_1": "garage heater", "unit_2": "bedroom lamp", "unit_3": "kitchen speaker", "unit_4": "living room heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_3", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 1, "unit_4": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"kitchen\",\"speaker\",true,2]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_0": 0, "unit_4": 0, "unit_2": 0, "unit_3": 1}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:49", "state_id": "smart_home_v2:test:49", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the bathroom lamp set to on; access=no; occupants=3; clock=19:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom speaker", "unit_1": "attic speaker", "unit_2": "dining room fan", "unit_3": "bathroom lamp", "unit_4": "hall lamp", "unit_5": "attic lamp", "unit_6": "hall speaker", "unit_7": "balcony heater", "unit_8": "living room lamp", "unit_9": "office speaker", "unit_10": "garden heater", "unit_11": "garden fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_3", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 1, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bathroom\",\"lamp\",false,3]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_3": 1, "unit_8": 0, "unit_1": 0, "unit_11": 0, "unit_7": 0, "unit_2": 0, "unit_0": 0, "unit_9": 0, "unit_10": 0, "unit_6": 0, "unit_5": 0, "unit_4": 0}, "execute": {"false": 0.96, "true": 0.04}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:50", "state_id": "smart_home_v2:test:50", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the hall heater set to off; access=no; occupants=0; clock=2:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "hall heater", "unit_1": "garage speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"hall\",\"heater\",false,0]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:52", "state_id": "smart_home_v2:test:52", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the balcony lamp set to on; access=no; occupants=2; clock=19:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom lamp", "unit_1": "living room lamp", "unit_2": "balcony lamp", "unit_3": "dining room speaker", "unit_4": "garden speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1, "unit_3": 0, "unit_4": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"balcony\",\"lamp\",false,2]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_2": 1, "unit_3": 0, "unit_4": 0, "unit_1": 0}, "execute": {"false": 0.96, "true": 0.04}, "risk": {"0": 0.02, "1": 0, "2": 0.98}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:54", "state_id": "smart_home_v2:test:54", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the garage heater set to off; access=yes; occupants=1; clock=11:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "office heater", "unit_1": "hall lamp", "unit_2": "living room fan", "unit_3": "office speaker", "unit_4": "garage heater", "unit_5": "garage fan", "unit_6": "attic speaker", "unit_7": "office lamp", "unit_8": "garden fan", "unit_9": "library fan", "unit_10": "attic fan", "unit_11": "bathroom lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_4", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 1, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garage\",\"heater\",true,1]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_5": 0, "unit_10": 0, "unit_4": 1, "unit_11": 0, "unit_3": 0, "unit_2": 0, "unit_7": 0, "unit_0": 0, "unit_9": 0, "unit_6": 0, "unit_1": 0, "unit_8": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:53", "state_id": "smart_home_v2:test:53", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the bathroom fan set to off; access=yes; occupants=1; clock=18:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "kitchen heater", "unit_1": "dining room heater", "unit_2": "bedroom speaker", "unit_3": "bathroom fan", "unit_4": "office fan", "unit_5": "balcony speaker", "unit_6": "bedroom lamp", "unit_7": "bedroom heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_3", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 1, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bathroom\",\"fan\",true,1]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_7": 0, "unit_5": 0, "unit_6": 0, "unit_4": 0, "unit_3": 1}, "execute": {"false": 0.06999999999999995, "true": 0.93}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:56", "state_id": "smart_home_v2:test:56", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the attic fan set to off; access=yes; occupants=1; clock=13:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "attic fan", "unit_1": "living room heater", "unit_2": "garden heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"attic\",\"fan\",true,1]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_2": 0, "unit_1": 0}, "execute": {"false": 0.050000000000000044, "true": 0.95}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:57", "state_id": "smart_home_v2:test:57", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the bedroom heater set to on; access=yes; occupants=2; clock=18:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garden fan", "unit_1": "living room speaker", "unit_2": "office fan", "unit_3": "bathroom speaker", "unit_4": "bedroom heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_4", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bedroom\",\"heater\",true,2]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_3": 0, "unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_4": 1}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:58", "state_id": "smart_home_v2:test:58", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the library speaker set to on; access=no; occupants=0; clock=10:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "hall lamp", "unit_1": "hall fan", "unit_2": "dining room fan", "unit_3": "kitchen heater", "unit_4": "library speaker", "unit_5": "garage lamp", "unit_6": "bedroom lamp", "unit_7": "bedroom speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_4", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 1, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"library\",\"speaker\",false,0]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_4": 1, "unit_3": 0, "unit_7": 0, "unit_1": 0, "unit_6": 0, "unit_0": 0, "unit_5": 0}, "execute": {"false": 0.95, "true": 0.05}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:59", "state_id": "smart_home_v2:test:59", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the library lamp set to off; access=no; occupants=0; clock=16:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garden heater", "unit_1": "garden fan", "unit_2": "library fan", "unit_3": "attic speaker", "unit_4": "bedroom speaker", "unit_5": "attic heater", "unit_6": "dining room lamp", "unit_7": "living room heater", "unit_8": "bedroom fan", "unit_9": "balcony fan", "unit_10": "bathroom heater", "unit_11": "library lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_11", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"library\",\"lamp\",false,0]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_8": 0, "unit_2": 0, "unit_5": 0, "unit_4": 0, "unit_0": 0, "unit_11": 1, "unit_7": 0, "unit_6": 0, "unit_9": 0, "unit_10": 0, "unit_3": 0}, "execute": {"false": 0.95, "true": 0.05}, "risk": {"0": 0.02, "1": 0, "2": 0.98}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:60", "state_id": "smart_home_v2:test:60", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the balcony fan set to on; access=no; occupants=3; clock=10:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garage speaker", "unit_1": "balcony fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"balcony\",\"fan\",false,3]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 0.95, "true": 0.05}, "risk": {"0": 0.02, "1": 0, "2": 0.98}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:61", "state_id": "smart_home_v2:test:61", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the balcony lamp set to off; access=yes; occupants=1; clock=7:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony lamp", "unit_1": "bathroom heater", "unit_2": "garden fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"balcony\",\"lamp\",true,1]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 0.040000000000000036, "true": 0.96}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:62", "state_id": "smart_home_v2:test:62", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the attic fan set to off; access=no; occupants=2; clock=19:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony speaker", "unit_1": "kitchen lamp", "unit_2": "attic fan", "unit_3": "library speaker", "unit_4": "bathroom heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1, "unit_3": 0, "unit_4": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"attic\",\"fan\",false,2]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_4": 0, "unit_2": 1, "unit_1": 0, "unit_3": 0, "unit_0": 0}, "execute": {"false": 0.95, "true": 0.05}, "risk": {"0": 0.06, "1": 0, "2": 0.94}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:63", "state_id": "smart_home_v2:test:63", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the library fan set to on; access=no; occupants=0; clock=6:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom speaker", "unit_1": "hall speaker", "unit_2": "garden speaker", "unit_3": "living room fan", "unit_4": "bathroom lamp", "unit_5": "bedroom heater", "unit_6": "balcony fan", "unit_7": "library fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_7", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"library\",\"fan\",false,0]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_3": 0, "unit_6": 0, "unit_0": 0, "unit_5": 0, "unit_4": 0, "unit_1": 0, "unit_7": 1}, "execute": {"false": 0.94, "true": 0.06}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:55", "state_id": "smart_home_v2:test:55", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the garage speaker set to on; access=no; occupants=0; clock=12:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garage speaker", "unit_1": "bathroom lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garage\",\"speaker\",false,0]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_0": 1}, "execute": {"false": 0.96, "true": 0.04}, "risk": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:0", "state_id": "catalog_lookup_v2:test:0", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00154; color=black; size=medium; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "black, medium", "candidate_1": "green, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00154\",\"black\",\"medium\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_0": 1}, "ready": {"false": 0.94, "true": 0.06}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:1", "state_id": "catalog_lookup_v2:test:1", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00124; color=white; size=small; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, medium", "candidate_1": "brown, small", "candidate_2": "white, small", "candidate_3": "gray, small", "candidate_4": "green, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_2", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 1, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00124\",\"white\",\"small\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_3": 0, "candidate_0": 0, "candidate_1": 0, "candidate_4": 0, "candidate_2": 1}, "ready": {"false": 0.18000000000000005, "true": 0.82}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:2", "state_id": "catalog_lookup_v2:test:2", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00028; color=blue; size=small; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "gray, medium", "candidate_1": "orange, small", "candidate_2": "white, large", "candidate_3": "pink, large", "candidate_4": "blue, small", "candidate_5": "yellow, small", "candidate_6": "red, large", "candidate_7": "gray, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_4", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 1, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00028\",\"blue\",\"small\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_3": 0, "candidate_2": 0, "candidate_6": 0, "candidate_7": 0, "candidate_4": 1, "candidate_0": 0, "candidate_5": 0}, "ready": {"false": 0.18000000000000005, "true": 0.82}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:3", "state_id": "catalog_lookup_v2:test:3", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00218; color=brown; size=small; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "cyan, large", "candidate_1": "pink, small", "candidate_2": "blue, small", "candidate_3": "gray, medium", "candidate_4": "pink, large", "candidate_5": "gray, small", "candidate_6": "orange, large", "candidate_7": "green, medium", "candidate_8": "brown, small", "candidate_9": "purple, medium", "candidate_10": "blue, medium", "candidate_11": "purple, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_8", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 1, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00218\",\"brown\",\"small\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_9": 0, "candidate_10": 0, "candidate_6": 0, "candidate_1": 0, "candidate_0": 0, "candidate_2": 0, "candidate_5": 0, "candidate_8": 1, "candidate_3": 0, "candidate_4": 0, "candidate_7": 0, "candidate_11": 0}, "ready": {"false": 0.9299999999999999, "true": 0.07}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:6", "state_id": "catalog_lookup_v2:test:6", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00258; color=gray; size=large; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "gray, large", "candidate_1": "gray, small", "candidate_2": "orange, large", "candidate_3": "cyan, medium", "candidate_4": "pink, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00258\",\"gray\",\"large\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_4": 0, "candidate_0": 1, "candidate_1": 0, "candidate_3": 0, "candidate_2": 0}, "ready": {"false": 0.9299999999999999, "true": 0.07}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:4", "state_id": "catalog_lookup_v2:test:4", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00158; color=black; size=medium; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "orange, large", "candidate_1": "cyan, small", "candidate_2": "brown, small", "candidate_3": "purple, small", "candidate_4": "white, medium", "candidate_5": "green, small", "candidate_6": "purple, large", "candidate_7": "red, small", "candidate_8": "blue, small", "candidate_9": "purple, medium", "candidate_10": "pink, small", "candidate_11": "black, large", "candidate_12": "gray, medium", "candidate_13": "pink, large", "candidate_14": "cyan, large", "candidate_15": "brown, large", "candidate_16": "red, large", "candidate_17": "gray, small", "candidate_18": "black, medium", "candidate_19": "cyan, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_18", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 1, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00158\",\"black\",\"medium\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_16": 0, "candidate_12": 0, "candidate_19": 0, "candidate_18": 1, "candidate_1": 0, "candidate_13": 0, "candidate_6": 0, "candidate_14": 0, "candidate_17": 0, "candidate_9": 0, "candidate_8": 0, "candidate_7": 0, "candidate_11": 0, "candidate_4": 0, "candidate_0": 0, "candidate_2": 0, "candidate_10": 0, "candidate_15": 0, "candidate_3": 0, "candidate_5": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:5", "state_id": "catalog_lookup_v2:test:5", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00256; color=gray; size=large; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "yellow, medium", "candidate_1": "gray, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00256\",\"gray\",\"large\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 0.8200000000000001, "true": 0.18}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:7", "state_id": "catalog_lookup_v2:test:7", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00030; color=blue; size=small; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "blue, large", "candidate_1": "orange, large", "candidate_2": "pink, medium", "candidate_3": "purple, large", "candidate_4": "green, large", "candidate_5": "black, medium", "candidate_6": "gray, large", "candidate_7": "blue, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_7", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00030\",\"blue\",\"small\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_4": 0, "candidate_1": 0, "candidate_6": 0, "candidate_0": 0, "candidate_2": 0, "candidate_7": 1, "candidate_3": 0, "candidate_5": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:9", "state_id": "catalog_lookup_v2:test:9", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00276; color=cyan; size=medium; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "red, large", "candidate_1": "cyan, large", "candidate_2": "brown, large", "candidate_3": "cyan, small", "candidate_4": "orange, medium", "candidate_5": "blue, large", "candidate_6": "green, large", "candidate_7": "yellow, large", "candidate_8": "pink, large", "candidate_9": "purple, large", "candidate_10": "pink, small", "candidate_11": "white, medium", "candidate_12": "brown, medium", "candidate_13": "orange, small", "candidate_14": "purple, medium", "candidate_15": "pink, medium", "candidate_16": "cyan, medium", "candidate_17": "black, large", "candidate_18": "red, small", "candidate_19": "blue, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_16", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 1, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00276\",\"cyan\",\"medium\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_14": 0, "candidate_15": 0, "candidate_16": 1, "candidate_13": 0, "candidate_18": 0, "candidate_7": 0, "candidate_0": 0, "candidate_10": 0, "candidate_11": 0, "candidate_9": 0, "candidate_2": 0, "candidate_5": 0, "candidate_12": 0, "candidate_4": 0, "candidate_1": 0, "candidate_19": 0, "candidate_6": 0, "candidate_17": 0, "candidate_8": 0, "candidate_3": 0}, "ready": {"false": 0.18000000000000005, "true": 0.82}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:8", "state_id": "catalog_lookup_v2:test:8", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00164; color=black; size=large; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "black, small", "candidate_1": "purple, medium", "candidate_2": "black, large", "candidate_3": "blue, medium", "candidate_4": "brown, small", "candidate_5": "pink, medium", "candidate_6": "green, small", "candidate_7": "yellow, medium", "candidate_8": "pink, small", "candidate_9": "green, medium", "candidate_10": "green, large", "candidate_11": "blue, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_2", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 1, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00164\",\"black\",\"large\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_5": 0, "candidate_0": 0, "candidate_2": 1, "candidate_10": 0, "candidate_9": 0, "candidate_4": 0, "candidate_8": 0, "candidate_11": 0, "candidate_1": 0, "candidate_3": 0, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 0.19999999999999996, "true": 0.8}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:10", "state_id": "catalog_lookup_v2:test:10", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00260; color=gray; size=large; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "brown, small", "candidate_1": "gray, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00260\",\"gray\",\"large\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 0.18999999999999995, "true": 0.81}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:12", "state_id": "catalog_lookup_v2:test:12", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00278; color=cyan; size=medium; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "black, large", "candidate_1": "blue, large", "candidate_2": "cyan, medium", "candidate_3": "gray, medium", "candidate_4": "purple, large", "candidate_5": "purple, medium", "candidate_6": "gray, small", "candidate_7": "green, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_2", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 1, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00278\",\"cyan\",\"medium\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_4": 0, "candidate_5": 0, "candidate_2": 1, "candidate_3": 0, "candidate_7": 0, "candidate_0": 0, "candidate_1": 0, "candidate_6": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:test:51", "state_id": "smart_home_v2:test:51", "family_id": "smart_home_v2", "split": "test", "state": "Home log: user wants the bedroom fan set to off; access=no; occupants=3; clock=4:00.", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom fan", "unit_1": "living room heater", "unit_2": "garage fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bedroom\",\"fan\",false,3]", "template_id": "smart_home_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_1": 0, "unit_0": 1}, "execute": {"false": 0.96, "true": 0.04}, "risk": {"0": 0.02, "1": 0, "2": 0.98}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:15", "state_id": "catalog_lookup_v2:test:15", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00173; color=orange; size=small; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "blue, large", "candidate_1": "orange, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00173\",\"orange\",\"small\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 0.15000000000000002, "true": 0.85}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:18", "state_id": "catalog_lookup_v2:test:18", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00106; color=purple; size=medium; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "cyan, large", "candidate_1": "green, large", "candidate_2": "gray, medium", "candidate_3": "white, large", "candidate_4": "red, small", "candidate_5": "orange, large", "candidate_6": "brown, medium", "candidate_7": "purple, medium", "candidate_8": "orange, medium", "candidate_9": "purple, large", "candidate_10": "blue, small", "candidate_11": "yellow, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_7", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 1, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00106\",\"purple\",\"medium\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_7": 1, "candidate_5": 0, "candidate_3": 0, "candidate_9": 0, "candidate_10": 0, "candidate_8": 0, "candidate_6": 0, "candidate_4": 0, "candidate_11": 0, "candidate_2": 0, "candidate_0": 0, "candidate_1": 0}, "ready": {"false": 0.95, "true": 0.05}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:16", "state_id": "catalog_lookup_v2:test:16", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00087; color=yellow; size=medium; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "pink, medium", "candidate_1": "yellow, medium", "candidate_2": "cyan, medium", "candidate_3": "cyan, large", "candidate_4": "purple, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00087\",\"yellow\",\"medium\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_3": 0, "candidate_2": 0, "candidate_4": 0, "candidate_0": 0, "candidate_1": 1}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:17", "state_id": "catalog_lookup_v2:test:17", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00031; color=blue; size=small; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "blue, large", "candidate_1": "green, small", "candidate_2": "red, large", "candidate_3": "yellow, large", "candidate_4": "blue, small", "candidate_5": "orange, large", "candidate_6": "purple, large", "candidate_7": "white, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_4", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 1, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00031\",\"blue\",\"small\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_2": 0, "candidate_4": 1, "candidate_5": 0, "candidate_0": 0, "candidate_6": 0, "candidate_1": 0, "candidate_3": 0, "candidate_7": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:19", "state_id": "catalog_lookup_v2:test:19", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00001; color=red; size=small; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "cyan, small", "candidate_1": "white, large", "candidate_2": "pink, small", "candidate_3": "brown, medium", "candidate_4": "pink, medium", "candidate_5": "gray, large", "candidate_6": "orange, medium", "candidate_7": "yellow, medium", "candidate_8": "purple, small", "candidate_9": "white, small", "candidate_10": "yellow, small", "candidate_11": "black, small", "candidate_12": "red, small", "candidate_13": "purple, medium", "candidate_14": "pink, large", "candidate_15": "white, medium", "candidate_16": "green, small", "candidate_17": "gray, medium", "candidate_18": "yellow, large", "candidate_19": "brown, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_12", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 1, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00001\",\"red\",\"small\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_16": 0, "candidate_8": 0, "candidate_4": 0, "candidate_11": 0, "candidate_0": 0, "candidate_1": 0, "candidate_6": 0, "candidate_10": 0, "candidate_13": 0, "candidate_5": 0, "candidate_9": 0, "candidate_14": 0, "candidate_3": 0, "candidate_7": 0, "candidate_12": 1, "candidate_2": 0, "candidate_15": 0, "candidate_19": 0, "candidate_17": 0, "candidate_18": 0}, "ready": {"false": 0.86, "true": 0.14}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:13", "state_id": "catalog_lookup_v2:test:13", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00080; color=yellow; size=medium; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "cyan, small", "candidate_1": "brown, medium", "candidate_2": "orange, large", "candidate_3": "black, large", "candidate_4": "white, large", "candidate_5": "orange, small", "candidate_6": "red, large", "candidate_7": "brown, small", "candidate_8": "green, large", "candidate_9": "green, small", "candidate_10": "yellow, medium", "candidate_11": "pink, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_10", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 1, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00080\",\"yellow\",\"medium\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_9": 0, "candidate_5": 0, "candidate_4": 0, "candidate_3": 0, "candidate_8": 0, "candidate_0": 0, "candidate_1": 0, "candidate_7": 0, "candidate_11": 0, "candidate_6": 0, "candidate_10": 1, "candidate_2": 0}, "ready": {"false": 0.83, "true": 0.17}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:20", "state_id": "catalog_lookup_v2:test:20", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00272; color=cyan; size=medium; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "cyan, medium", "candidate_1": "purple, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00272\",\"cyan\",\"medium\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_0": 1}, "ready": {"false": 0.8200000000000001, "true": 0.18}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:11", "state_id": "catalog_lookup_v2:test:11", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00118; color=purple; size=large; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "black, small", "candidate_1": "purple, large", "candidate_2": "gray, medium", "candidate_3": "pink, small", "candidate_4": "brown, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00118\",\"purple\",\"large\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 1, "candidate_0": 0, "candidate_2": 0, "candidate_4": 0, "candidate_3": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:21", "state_id": "catalog_lookup_v2:test:21", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00088; color=yellow; size=large; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "pink, medium", "candidate_1": "orange, small", "candidate_2": "cyan, medium", "candidate_3": "blue, small", "candidate_4": "yellow, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_4", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00088\",\"yellow\",\"large\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_3": 0, "candidate_4": 1, "candidate_1": 0, "candidate_2": 0, "candidate_0": 0}, "ready": {"false": 0.8200000000000001, "true": 0.18}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:22", "state_id": "catalog_lookup_v2:test:22", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00284; color=cyan; size=large; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "cyan, small", "candidate_1": "white, medium", "candidate_2": "pink, large", "candidate_3": "black, large", "candidate_4": "cyan, large", "candidate_5": "red, medium", "candidate_6": "white, small", "candidate_7": "brown, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_4", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 1, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00284\",\"cyan\",\"large\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_3": 0, "candidate_2": 0, "candidate_6": 0, "candidate_7": 0, "candidate_4": 1, "candidate_0": 0, "candidate_5": 0}, "ready": {"false": 0.18999999999999995, "true": 0.81}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:14", "state_id": "catalog_lookup_v2:test:14", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00081; color=yellow; size=medium; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "yellow, medium", "candidate_1": "gray, small", "candidate_2": "blue, medium", "candidate_3": "green, large", "candidate_4": "yellow, small", "candidate_5": "white, small", "candidate_6": "pink, small", "candidate_7": "brown, medium", "candidate_8": "pink, large", "candidate_9": "black, large", "candidate_10": "red, medium", "candidate_11": "cyan, large", "candidate_12": "green, medium", "candidate_13": "green, small", "candidate_14": "purple, small", "candidate_15": "brown, large", "candidate_16": "orange, large", "candidate_17": "gray, large", "candidate_18": "orange, medium", "candidate_19": "cyan, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00081\",\"yellow\",\"medium\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_9": 0, "candidate_5": 0, "candidate_0": 1, "candidate_14": 0, "candidate_8": 0, "candidate_1": 0, "candidate_18": 0, "candidate_11": 0, "candidate_12": 0, "candidate_10": 0, "candidate_2": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_3": 0, "candidate_4": 0, "candidate_19": 0, "candidate_13": 0, "candidate_7": 0, "candidate_6": 0}, "ready": {"false": 0.84, "true": 0.16}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:23", "state_id": "catalog_lookup_v2:test:23", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00054; color=green; size=small; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "black, medium", "candidate_1": "purple, medium", "candidate_2": "black, large", "candidate_3": "gray, large", "candidate_4": "yellow, small", "candidate_5": "pink, small", "candidate_6": "brown, small", "candidate_7": "gray, small", "candidate_8": "green, small", "candidate_9": "orange, medium", "candidate_10": "cyan, medium", "candidate_11": "cyan, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_8", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 1, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00054\",\"green\",\"small\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_5": 0, "candidate_6": 0, "candidate_3": 0, "candidate_8": 1, "candidate_7": 0, "candidate_2": 0, "candidate_4": 0, "candidate_10": 0, "candidate_0": 0, "candidate_9": 0, "candidate_11": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:25", "state_id": "catalog_lookup_v2:test:25", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00234; color=brown; size=large; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "brown, large", "candidate_1": "orange, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00234\",\"brown\",\"large\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_0": 1}, "ready": {"false": 0.9299999999999999, "true": 0.07}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:26", "state_id": "catalog_lookup_v2:test:26", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00091; color=yellow; size=large; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "red, small", "candidate_1": "cyan, medium", "candidate_2": "yellow, large", "candidate_3": "gray, large", "candidate_4": "brown, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_2", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 1, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00091\",\"yellow\",\"large\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_4": 0, "candidate_0": 0, "candidate_3": 0, "candidate_2": 1}, "ready": {"false": 0.95, "true": 0.05}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:24", "state_id": "catalog_lookup_v2:test:24", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00029; color=blue; size=small; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "yellow, medium", "candidate_1": "red, medium", "candidate_2": "cyan, small", "candidate_3": "white, medium", "candidate_4": "gray, medium", "candidate_5": "pink, small", "candidate_6": "blue, large", "candidate_7": "white, small", "candidate_8": "pink, large", "candidate_9": "orange, medium", "candidate_10": "purple, medium", "candidate_11": "red, small", "candidate_12": "blue, medium", "candidate_13": "red, large", "candidate_14": "green, small", "candidate_15": "brown, small", "candidate_16": "yellow, small", "candidate_17": "blue, small", "candidate_18": "orange, small", "candidate_19": "brown, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_17", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 1, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00029\",\"blue\",\"small\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_18": 0, "candidate_5": 0, "candidate_12": 0, "candidate_15": 0, "candidate_7": 0, "candidate_4": 0, "candidate_3": 0, "candidate_6": 0, "candidate_14": 0, "candidate_2": 0, "candidate_17": 1, "candidate_11": 0, "candidate_13": 0, "candidate_1": 0, "candidate_16": 0, "candidate_8": 0, "candidate_10": 0, "candidate_0": 0, "candidate_19": 0, "candidate_9": 0}, "ready": {"false": 0.14, "true": 0.86}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:28", "state_id": "catalog_lookup_v2:test:28", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00269; color=cyan; size=small; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "gray, small", "candidate_1": "pink, large", "candidate_2": "black, medium", "candidate_3": "white, medium", "candidate_4": "white, small", "candidate_5": "brown, large", "candidate_6": "black, large", "candidate_7": "blue, medium", "candidate_8": "green, small", "candidate_9": "yellow, large", "candidate_10": "cyan, small", "candidate_11": "black, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_10", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 1, "candidate_11": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00269\",\"cyan\",\"small\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_4": 0, "candidate_11": 0, "candidate_3": 0, "candidate_10": 1, "candidate_0": 0, "candidate_9": 0, "candidate_8": 0, "candidate_6": 0, "candidate_2": 0, "candidate_1": 0, "candidate_7": 0, "candidate_5": 0}, "ready": {"false": 0.15000000000000002, "true": 0.85}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:30", "state_id": "catalog_lookup_v2:test:30", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00177; color=orange; size=medium; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, large", "candidate_1": "orange, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00177\",\"orange\",\"medium\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 0.87, "true": 0.13}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:31", "state_id": "catalog_lookup_v2:test:31", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00099; color=purple; size=small; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "yellow, medium", "candidate_1": "purple, medium", "candidate_2": "black, large", "candidate_3": "green, medium", "candidate_4": "purple, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_4", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00099\",\"purple\",\"small\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_4": 1, "candidate_3": 0, "candidate_0": 0, "candidate_2": 0, "candidate_1": 0}, "ready": {"false": 0.9299999999999999, "true": 0.07}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:32", "state_id": "catalog_lookup_v2:test:32", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00041; color=blue; size=large; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "brown, large", "candidate_1": "gray, large", "candidate_2": "red, small", "candidate_3": "white, small", "candidate_4": "pink, small", "candidate_5": "blue, large", "candidate_6": "purple, medium", "candidate_7": "yellow, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_5", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 1, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00041\",\"blue\",\"large\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_5": 1, "candidate_4": 0, "candidate_7": 0, "candidate_1": 0, "candidate_0": 0, "candidate_3": 0, "candidate_6": 0, "candidate_2": 0}, "ready": {"false": 0.86, "true": 0.14}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:27", "state_id": "catalog_lookup_v2:test:27", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00006; color=red; size=small; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "red, large", "candidate_1": "blue, large", "candidate_2": "orange, small", "candidate_3": "yellow, small", "candidate_4": "yellow, large", "candidate_5": "gray, medium", "candidate_6": "black, medium", "candidate_7": "red, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_7", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00006\",\"red\",\"small\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_3": 0, "candidate_0": 0, "candidate_6": 0, "candidate_4": 0, "candidate_1": 0, "candidate_5": 0, "candidate_2": 0, "candidate_7": 1}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:33", "state_id": "catalog_lookup_v2:test:33", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00286; color=cyan; size=large; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "gray, large", "candidate_1": "white, medium", "candidate_2": "cyan, medium", "candidate_3": "red, small", "candidate_4": "red, large", "candidate_5": "pink, medium", "candidate_6": "green, large", "candidate_7": "gray, medium", "candidate_8": "green, medium", "candidate_9": "yellow, medium", "candidate_10": "cyan, large", "candidate_11": "brown, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_10", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 1, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00286\",\"cyan\",\"large\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_4": 0, "candidate_11": 0, "candidate_10": 1, "candidate_6": 0, "candidate_3": 0, "candidate_5": 0, "candidate_0": 0, "candidate_9": 0, "candidate_8": 0, "candidate_7": 0, "candidate_2": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:35", "state_id": "catalog_lookup_v2:test:35", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00244; color=gray; size=small; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "yellow, large", "candidate_1": "gray, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00244\",\"gray\",\"small\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 0.20999999999999996, "true": 0.79}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:29", "state_id": "catalog_lookup_v2:test:29", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00262; color=gray; size=large; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "brown, medium", "candidate_1": "gray, small", "candidate_2": "red, medium", "candidate_3": "green, medium", "candidate_4": "red, large", "candidate_5": "white, medium", "candidate_6": "white, large", "candidate_7": "green, large", "candidate_8": "gray, large", "candidate_9": "gray, medium", "candidate_10": "orange, small", "candidate_11": "cyan, large", "candidate_12": "pink, small", "candidate_13": "black, medium", "candidate_14": "blue, small", "candidate_15": "green, small", "candidate_16": "purple, medium", "candidate_17": "orange, medium", "candidate_18": "yellow, small", "candidate_19": "pink, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_8", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 1, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00262\",\"gray\",\"large\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_17": 0, "candidate_2": 0, "candidate_4": 0, "candidate_6": 0, "candidate_12": 0, "candidate_8": 1, "candidate_0": 0, "candidate_10": 0, "candidate_9": 0, "candidate_18": 0, "candidate_5": 0, "candidate_14": 0, "candidate_16": 0, "candidate_13": 0, "candidate_19": 0, "candidate_15": 0, "candidate_7": 0, "candidate_11": 0, "candidate_3": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:34", "state_id": "catalog_lookup_v2:test:34", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00140; color=white; size=large; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "blue, large", "candidate_1": "pink, small", "candidate_2": "pink, medium", "candidate_3": "cyan, small", "candidate_4": "cyan, medium", "candidate_5": "white, small", "candidate_6": "red, medium", "candidate_7": "yellow, small", "candidate_8": "purple, small", "candidate_9": "purple, large", "candidate_10": "brown, medium", "candidate_11": "red, small", "candidate_12": "gray, small", "candidate_13": "brown, small", "candidate_14": "white, large", "candidate_15": "pink, large", "candidate_16": "red, large", "candidate_17": "orange, large", "candidate_18": "brown, large", "candidate_19": "orange, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_14", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 1, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00140\",\"white\",\"large\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_19": 0, "candidate_13": 0, "candidate_12": 0, "candidate_8": 0, "candidate_17": 0, "candidate_16": 0, "candidate_4": 0, "candidate_10": 0, "candidate_3": 0, "candidate_1": 0, "candidate_11": 0, "candidate_14": 1, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_0": 0, "candidate_15": 0, "candidate_2": 0, "candidate_18": 0, "candidate_9": 0}, "ready": {"false": 0.19999999999999996, "true": 0.8}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:36", "state_id": "catalog_lookup_v2:test:36", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00162; color=black; size=large; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "blue, small", "candidate_1": "purple, small", "candidate_2": "black, large", "candidate_3": "blue, large", "candidate_4": "blue, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_2", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 1, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00162\",\"black\",\"large\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_4": 0, "candidate_1": 0, "candidate_0": 0, "candidate_3": 0, "candidate_2": 1}, "ready": {"false": 0.92, "true": 0.08}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:37", "state_id": "catalog_lookup_v2:test:37", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00161; color=black; size=large; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "pink, medium", "candidate_1": "cyan, medium", "candidate_2": "purple, large", "candidate_3": "black, large", "candidate_4": "white, small", "candidate_5": "orange, small", "candidate_6": "blue, small", "candidate_7": "blue, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_3", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 1, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00161\",\"black\",\"large\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_7": 0, "candidate_6": 0, "candidate_2": 0, "candidate_5": 0, "candidate_3": 1, "candidate_4": 0}, "ready": {"false": 0.86, "true": 0.14}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:39", "state_id": "catalog_lookup_v2:test:39", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00078; color=yellow; size=small; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "yellow, medium", "candidate_1": "gray, medium", "candidate_2": "white, medium", "candidate_3": "pink, large", "candidate_4": "brown, small", "candidate_5": "black, large", "candidate_6": "orange, large", "candidate_7": "white, small", "candidate_8": "pink, medium", "candidate_9": "purple, small", "candidate_10": "gray, small", "candidate_11": "blue, large", "candidate_12": "green, small", "candidate_13": "purple, medium", "candidate_14": "white, large", "candidate_15": "red, medium", "candidate_16": "orange, small", "candidate_17": "green, medium", "candidate_18": "cyan, medium", "candidate_19": "yellow, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_19", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00078\",\"yellow\",\"small\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_12": 0, "candidate_19": 1, "candidate_1": 0, "candidate_6": 0, "candidate_2": 0, "candidate_18": 0, "candidate_17": 0, "candidate_5": 0, "candidate_13": 0, "candidate_15": 0, "candidate_7": 0, "candidate_9": 0, "candidate_10": 0, "candidate_0": 0, "candidate_16": 0, "candidate_14": 0, "candidate_8": 0, "candidate_3": 0, "candidate_4": 0, "candidate_11": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:38", "state_id": "catalog_lookup_v2:test:38", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00137; color=white; size=large; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "orange, small", "candidate_1": "green, medium", "candidate_2": "brown, small", "candidate_3": "yellow, medium", "candidate_4": "pink, large", "candidate_5": "gray, small", "candidate_6": "black, small", "candidate_7": "white, large", "candidate_8": "black, medium", "candidate_9": "green, large", "candidate_10": "red, small", "candidate_11": "blue, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_7", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 1, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00137\",\"white\",\"large\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_3": 0, "candidate_9": 0, "candidate_6": 0, "candidate_1": 0, "candidate_5": 0, "candidate_10": 0, "candidate_0": 0, "candidate_4": 0, "candidate_8": 0, "candidate_11": 0, "candidate_7": 1, "candidate_2": 0}, "ready": {"false": 0.86, "true": 0.14}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:41", "state_id": "catalog_lookup_v2:test:41", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00049; color=green; size=small; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "brown, small", "candidate_1": "white, large", "candidate_2": "white, small", "candidate_3": "black, small", "candidate_4": "green, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_4", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00049\",\"green\",\"small\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_3": 0, "candidate_2": 0, "candidate_0": 0, "candidate_4": 1}, "ready": {"false": 0.85, "true": 0.15}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:40", "state_id": "catalog_lookup_v2:test:40", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00108; color=purple; size=medium; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "green, medium", "candidate_1": "purple, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00108\",\"purple\",\"medium\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 1, "candidate_0": 0}, "ready": {"false": 0.16000000000000003, "true": 0.84}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:42", "state_id": "catalog_lookup_v2:test:42", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00083; color=yellow; size=medium; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "red, medium", "candidate_1": "orange, large", "candidate_2": "black, small", "candidate_3": "blue, small", "candidate_4": "purple, large", "candidate_5": "brown, small", "candidate_6": "cyan, small", "candidate_7": "yellow, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_7", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00083\",\"yellow\",\"medium\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_5": 0, "candidate_6": 0, "candidate_7": 1, "candidate_3": 0, "candidate_0": 0, "candidate_2": 0, "candidate_4": 0, "candidate_1": 0}, "ready": {"false": 0.95, "true": 0.05}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:44", "state_id": "catalog_lookup_v2:test:44", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00046; color=blue; size=large; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "green, large", "candidate_1": "blue, large", "candidate_2": "green, medium", "candidate_3": "pink, medium", "candidate_4": "yellow, large", "candidate_5": "white, small", "candidate_6": "white, large", "candidate_7": "brown, small", "candidate_8": "red, medium", "candidate_9": "red, large", "candidate_10": "cyan, large", "candidate_11": "orange, large", "candidate_12": "white, medium", "candidate_13": "blue, medium", "candidate_14": "orange, medium", "candidate_15": "yellow, medium", "candidate_16": "black, medium", "candidate_17": "black, small", "candidate_18": "purple, large", "candidate_19": "brown, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00046\",\"blue\",\"large\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_14": 0, "candidate_18": 0, "candidate_9": 0, "candidate_4": 0, "candidate_1": 1, "candidate_7": 0, "candidate_5": 0, "candidate_19": 0, "candidate_17": 0, "candidate_0": 0, "candidate_2": 0, "candidate_10": 0, "candidate_13": 0, "candidate_8": 0, "candidate_15": 0, "candidate_11": 0, "candidate_12": 0, "candidate_6": 0, "candidate_3": 0, "candidate_16": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:43", "state_id": "catalog_lookup_v2:test:43", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00181; color=orange; size=medium; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "black, medium", "candidate_1": "purple, medium", "candidate_2": "gray, medium", "candidate_3": "green, medium", "candidate_4": "yellow, large", "candidate_5": "orange, large", "candidate_6": "black, small", "candidate_7": "white, small", "candidate_8": "blue, large", "candidate_9": "pink, medium", "candidate_10": "orange, medium", "candidate_11": "green, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_10", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 1, "candidate_11": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00181\",\"orange\",\"medium\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_11": 0, "candidate_10": 1, "candidate_6": 0, "candidate_2": 0, "candidate_8": 0, "candidate_5": 0, "candidate_4": 0, "candidate_0": 0, "candidate_7": 0, "candidate_3": 0, "candidate_9": 0}, "ready": {"false": 0.14, "true": 0.86}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:46", "state_id": "catalog_lookup_v2:test:46", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00152; color=black; size=medium; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "brown, small", "candidate_1": "blue, medium", "candidate_2": "red, medium", "candidate_3": "black, medium", "candidate_4": "cyan, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_3", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 1, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00152\",\"black\",\"medium\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_4": 0, "candidate_0": 0, "candidate_3": 1, "candidate_2": 0}, "ready": {"false": 0.81, "true": 0.19}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:45", "state_id": "catalog_lookup_v2:test:45", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00094; color=yellow; size=large; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "yellow, large", "candidate_1": "black, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00094\",\"yellow\",\"large\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_0": 1}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:1", "state_id": "known_chance_v2:test:1", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: gray=5, cyan=7, white=9, pink=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"pink": "The sampled ball is pink", "gray": "The sampled ball is gray", "cyan": "The sampled ball is cyan", "white": "The sampled ball is white"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a white ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: gray, pink", "The drawn color is one of: cyan", "The drawn color is one of: white"]}}, "gold": {}, "gold_probs": {"outcome": {"pink": 0.2222222222222222, "gray": 0.18518518518518517, "cyan": 0.25925925925925924, "white": 0.3333333333333333}, "focus": {"false": 0.6666666666666667, "true": 0.3333333333333333}, "group": {"0": 0.4074074074074074, "1": 0.25925925925925924, "2": 0.3333333333333333}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"7/27\"],[\"gray\",\"5/27\"],[\"pink\",\"2/9\"],[\"white\",\"1/3\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.01, "cyan": 0.02, "white": 0.81, "gray": 0.15}, "focus": {"false": 0.72, "true": 0.28}, "group": {"0": 0.03, "1": 0.02, "2": 0.95}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:3", "state_id": "known_chance_v2:test:3", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: blue=2, orange=5, purple=6, green=4, brown=3, yellow=4, gray=8, cyan=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"orange": "The sampled ball is orange", "green": "The sampled ball is green", "brown": "The sampled ball is brown", "cyan": "The sampled ball is cyan", "gray": "The sampled ball is gray", "blue": "The sampled ball is blue", "purple": "The sampled ball is purple", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a purple ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue, green, gray", "The drawn color is one of: orange, brown, cyan", "The drawn color is one of: purple, yellow"]}}, "gold": {}, "gold_probs": {"outcome": {"orange": 0.1282051282051282, "green": 0.10256410256410256, "brown": 0.07692307692307693, "cyan": 0.1794871794871795, "gray": 0.20512820512820512, "blue": 0.05128205128205128, "purple": 0.15384615384615385, "yellow": 0.10256410256410256}, "focus": {"false": 0.8461538461538461, "true": 0.15384615384615385}, "group": {"0": 0.358974358974359, "1": 0.3846153846153846, "2": 0.2564102564102564}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"2/39\"],[\"brown\",\"1/13\"],[\"cyan\",\"7/39\"],[\"gray\",\"8/39\"],[\"green\",\"4/39\"],[\"orange\",\"5/39\"],[\"purple\",\"2/13\"],[\"yellow\",\"4/39\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0, "brown": 0, "purple": 0, "green": 0, "gray": 0.92, "blue": 0.08, "yellow": 0, "orange": 0}, "focus": {"false": 0.83, "true": 0.17}, "group": {"0": 0.39, "1": 0.47, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:0", "state_id": "known_chance_v2:test:0", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: white=1, red=2, black=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "white": "The sampled ball is white", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a white ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white", "The drawn color is one of: red", "The drawn color is one of: black"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.6666666666666666, "white": 0.1111111111111111, "red": 0.2222222222222222}, "focus": {"false": 0.8888888888888888, "true": 0.1111111111111111}, "group": {"0": 0.1111111111111111, "1": 0.2222222222222222, "2": 0.6666666666666666}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"2/3\"],[\"red\",\"2/9\"],[\"white\",\"1/9\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"red": 0.01, "black": 0.95, "white": 0.04}, "focus": {"false": 0.89, "true": 0.11}, "group": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:2", "state_id": "known_chance_v2:test:2", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: white=1, cyan=4, pink=9, green=1, brown=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "brown": "The sampled ball is brown", "pink": "The sampled ball is pink", "green": "The sampled ball is green", "white": "The sampled ball is white"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a green ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white, green", "The drawn color is one of: cyan, brown", "The drawn color is one of: pink"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.2222222222222222, "brown": 0.16666666666666666, "pink": 0.5, "green": 0.05555555555555555, "white": 0.05555555555555555}, "focus": {"false": 0.9444444444444444, "true": 0.05555555555555555}, "group": {"0": 0.1111111111111111, "1": 0.38888888888888884, "2": 0.5}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"1/6\"],[\"cyan\",\"2/9\"],[\"green\",\"1/18\"],[\"pink\",\"1/2\"],[\"white\",\"1/18\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0.01, "white": 0.02, "pink": 0.97, "brown": 0, "green": 0}, "focus": {"false": 0.9299999999999999, "true": 0.07}, "group": {"0": 0.01, "1": 0.06, "2": 0.93}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:5", "state_id": "known_chance_v2:test:5", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: blue=1, gray=8, black=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "blue": "The sampled ball is blue", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue", "The drawn color is one of: gray", "The drawn color is one of: black"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.1, "blue": 0.1, "gray": 0.8}, "focus": {"false": 0.9, "true": 0.1}, "group": {"0": 0.1, "1": 0.8, "2": 0.1}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/10\"],[\"blue\",\"1/10\"],[\"gray\",\"4/5\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"black": 0, "blue": 0.03, "gray": 0.97}, "focus": {"false": 0.92, "true": 0.08}, "group": {"0": 0, "1": 1, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:6", "state_id": "known_chance_v2:test:6", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: white=3, red=2, black=6, cyan=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "white": "The sampled ball is white", "black": "The sampled ball is black", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a white ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white, cyan", "The drawn color is one of: red", "The drawn color is one of: black"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.21428571428571427, "white": 0.21428571428571427, "black": 0.42857142857142855, "red": 0.14285714285714285}, "focus": {"false": 0.7857142857142857, "true": 0.21428571428571427}, "group": {"0": 0.42857142857142855, "1": 0.14285714285714285, "2": 0.42857142857142855}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"3/7\"],[\"cyan\",\"3/14\"],[\"red\",\"1/7\"],[\"white\",\"3/14\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"red": 0.01, "black": 0.93, "cyan": 0, "white": 0.06}, "focus": {"false": 0.8200000000000001, "true": 0.18}, "group": {"0": 0.03, "1": 0.01, "2": 0.96}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:7", "state_id": "known_chance_v2:test:7", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: yellow=2, green=8, cyan=1, blue=1, white=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "white": "The sampled ball is white", "blue": "The sampled ball is blue", "cyan": "The sampled ball is cyan", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a green ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: yellow, blue", "The drawn color is one of: green, white", "The drawn color is one of: cyan"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.6153846153846154, "white": 0.07692307692307693, "blue": 0.07692307692307693, "cyan": 0.07692307692307693, "yellow": 0.15384615384615385}, "focus": {"false": 0.3846153846153846, "true": 0.6153846153846154}, "group": {"0": 0.23076923076923078, "1": 0.6923076923076923, "2": 0.07692307692307693}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"1/13\"],[\"cyan\",\"1/13\"],[\"green\",\"8/13\"],[\"white\",\"1/13\"],[\"yellow\",\"2/13\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0, "white": 0, "yellow": 0.02, "green": 0.98, "blue": 0}, "focus": {"false": 0.47, "true": 0.53}, "group": {"0": 0.02, "1": 0.97, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:test:47", "state_id": "catalog_lookup_v2:test:47", "family_id": "catalog_lookup_v2", "split": "test", "state": "Independent stock entry: item=SKU-00283; color=cyan; size=large; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, large", "candidate_1": "white, small", "candidate_2": "gray, large", "candidate_3": "red, large", "candidate_4": "cyan, large", "candidate_5": "blue, small", "candidate_6": "yellow, large", "candidate_7": "black, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_4", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 1, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00283\",\"cyan\",\"large\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_4": 1, "candidate_1": 0, "candidate_6": 0, "candidate_0": 0, "candidate_2": 0, "candidate_7": 0, "candidate_3": 0, "candidate_5": 0}, "ready": {"false": 0.9299999999999999, "true": 0.07}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:8", "state_id": "known_chance_v2:test:8", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: cyan=3, purple=7, black=8, pink=1, gray=8, blue=4, green=6, white=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "pink": "The sampled ball is pink", "gray": "The sampled ball is gray", "green": "The sampled ball is green", "cyan": "The sampled ball is cyan", "black": "The sampled ball is black", "white": "The sampled ball is white", "purple": "The sampled ball is purple"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: cyan, pink, green", "The drawn color is one of: purple, gray, white", "The drawn color is one of: black, blue"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.09523809523809523, "pink": 0.023809523809523808, "gray": 0.19047619047619047, "green": 0.14285714285714285, "cyan": 0.07142857142857142, "black": 0.19047619047619047, "white": 0.11904761904761904, "purple": 0.16666666666666666}, "focus": {"false": 0.9285714285714286, "true": 0.07142857142857142}, "group": {"0": 0.23809523809523808, "1": 0.47619047619047616, "2": 0.2857142857142857}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"4/21\"],[\"blue\",\"2/21\"],[\"cyan\",\"1/14\"],[\"gray\",\"4/21\"],[\"green\",\"1/7\"],[\"pink\",\"1/42\"],[\"purple\",\"1/6\"],[\"white\",\"5/42\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.06, "black": 0.71, "cyan": 0.03, "purple": 0.08, "gray": 0.09, "green": 0.01, "blue": 0.01, "white": 0.01}, "focus": {"false": 0.9, "true": 0.1}, "group": {"0": 0.01, "1": 0.9, "2": 0.09}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:10", "state_id": "known_chance_v2:test:10", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: yellow=3, brown=4, pink=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"pink": "The sampled ball is pink", "yellow": "The sampled ball is yellow", "brown": "The sampled ball is brown"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: yellow", "The drawn color is one of: brown", "The drawn color is one of: pink"]}}, "gold": {}, "gold_probs": {"outcome": {"pink": 0.125, "yellow": 0.375, "brown": 0.5}, "focus": {"false": 0.625, "true": 0.375}, "group": {"0": 0.375, "1": 0.5, "2": 0.125}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"1/2\"],[\"pink\",\"1/8\"],[\"yellow\",\"3/8\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.02, "brown": 0.64, "yellow": 0.34}, "focus": {"false": 0.73, "true": 0.27}, "group": {"0": 0.04, "1": 0.95, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:11", "state_id": "known_chance_v2:test:11", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: orange=6, cyan=6, pink=5, red=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"red": "The sampled ball is red", "orange": "The sampled ball is orange", "cyan": "The sampled ball is cyan", "pink": "The sampled ball is pink"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: orange, red", "The drawn color is one of: cyan", "The drawn color is one of: pink"]}}, "gold": {}, "gold_probs": {"outcome": {"red": 0.15, "orange": 0.3, "cyan": 0.3, "pink": 0.25}, "focus": {"false": 0.7, "true": 0.3}, "group": {"0": 0.44999999999999996, "1": 0.3, "2": 0.25}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"3/10\"],[\"orange\",\"3/10\"],[\"pink\",\"1/4\"],[\"red\",\"3/20\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.03, "red": 0.02, "cyan": 0.05, "orange": 0.9}, "focus": {"false": 0.75, "true": 0.25}, "group": {"0": 0.79, "1": 0.16, "2": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:9", "state_id": "known_chance_v2:test:9", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: pink=6, cyan=7, gray=6, green=2, orange=4, white=6, red=8, blue=8, yellow=2, brown=2, purple=4, black=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "cyan": "The sampled ball is cyan", "white": "The sampled ball is white", "yellow": "The sampled ball is yellow", "orange": "The sampled ball is orange", "red": "The sampled ball is red", "purple": "The sampled ball is purple", "black": "The sampled ball is black", "gray": "The sampled ball is gray", "green": "The sampled ball is green", "pink": "The sampled ball is pink", "brown": "The sampled ball is brown"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a gray ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: pink, green, red, brown", "The drawn color is one of: cyan, orange, blue, purple", "The drawn color is one of: gray, white, yellow, black"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.14035087719298245, "cyan": 0.12280701754385964, "white": 0.10526315789473684, "yellow": 0.03508771929824561, "orange": 0.07017543859649122, "red": 0.14035087719298245, "purple": 0.07017543859649122, "black": 0.03508771929824561, "gray": 0.10526315789473684, "green": 0.03508771929824561, "pink": 0.10526315789473684, "brown": 0.03508771929824561}, "focus": {"false": 0.8947368421052632, "true": 0.10526315789473684}, "group": {"0": 0.3157894736842105, "1": 0.40350877192982454, "2": 0.2807017543859649}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"2/57\"],[\"blue\",\"8/57\"],[\"brown\",\"2/57\"],[\"cyan\",\"7/57\"],[\"gray\",\"2/19\"],[\"green\",\"2/57\"],[\"orange\",\"4/57\"],[\"pink\",\"2/19\"],[\"purple\",\"4/57\"],[\"red\",\"8/57\"],[\"white\",\"2/19\"],[\"yellow\",\"2/57\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0.01, "purple": 0.01, "yellow": 0.01, "pink": 0.13, "gray": 0.02, "red": 0.52, "black": 0.01, "cyan": 0.14, "orange": 0.01, "brown": 0, "blue": 0.13, "green": 0.01}, "focus": {"false": 0.85, "true": 0.15}, "group": {"0": 0.3, "1": 0.62, "2": 0.08}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:4", "state_id": "known_chance_v2:test:4", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: cyan=7, brown=8, purple=6, red=1, gray=3, pink=9, green=2, black=4, blue=7, yellow=4, orange=1, white=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "purple": "The sampled ball is purple", "black": "The sampled ball is black", "red": "The sampled ball is red", "brown": "The sampled ball is brown", "white": "The sampled ball is white", "gray": "The sampled ball is gray", "green": "The sampled ball is green", "blue": "The sampled ball is blue", "yellow": "The sampled ball is yellow", "orange": "The sampled ball is orange", "pink": "The sampled ball is pink"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a purple ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: cyan, red, green, yellow", "The drawn color is one of: brown, gray, black, orange", "The drawn color is one of: purple, pink, blue, white"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.11864406779661017, "purple": 0.1016949152542373, "black": 0.06779661016949153, "red": 0.01694915254237288, "brown": 0.13559322033898305, "white": 0.11864406779661017, "gray": 0.05084745762711865, "green": 0.03389830508474576, "blue": 0.11864406779661017, "yellow": 0.06779661016949153, "orange": 0.01694915254237288, "pink": 0.15254237288135594}, "focus": {"false": 0.8983050847457628, "true": 0.1016949152542373}, "group": {"0": 0.23728813559322035, "1": 0.2711864406779661, "2": 0.4915254237288136}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"4/59\"],[\"blue\",\"7/59\"],[\"brown\",\"8/59\"],[\"cyan\",\"7/59\"],[\"gray\",\"3/59\"],[\"green\",\"2/59\"],[\"orange\",\"1/59\"],[\"pink\",\"9/59\"],[\"purple\",\"6/59\"],[\"red\",\"1/59\"],[\"white\",\"7/59\"],[\"yellow\",\"4/59\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0, "gray": 0, "brown": 0.08, "orange": 0, "white": 0.02, "yellow": 0, "black": 0, "cyan": 0.02, "red": 0.01, "pink": 0.87, "purple": 0, "blue": 0}, "focus": {"false": 0.88, "true": 0.12}, "group": {"0": 0.02, "1": 0.03, "2": 0.95}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:12", "state_id": "known_chance_v2:test:12", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: brown=6, red=7, pink=1, black=2, orange=9. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"brown": "The sampled ball is brown", "orange": "The sampled ball is orange", "black": "The sampled ball is black", "pink": "The sampled ball is pink", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a pink ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: brown, black", "The drawn color is one of: red, orange", "The drawn color is one of: pink"]}}, "gold": {}, "gold_probs": {"outcome": {"brown": 0.24, "orange": 0.36, "black": 0.08, "pink": 0.04, "red": 0.28}, "focus": {"false": 0.96, "true": 0.04}, "group": {"0": 0.32, "1": 0.64, "2": 0.04}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"2/25\"],[\"brown\",\"6/25\"],[\"orange\",\"9/25\"],[\"pink\",\"1/25\"],[\"red\",\"7/25\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"orange": 0.9, "black": 0, "brown": 0.05, "pink": 0.02, "red": 0.03}, "focus": {"false": 0.95, "true": 0.05}, "group": {"0": 0.01, "1": 0.98, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:13", "state_id": "known_chance_v2:test:13", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: yellow=1, orange=8, pink=2, black=7, purple=3, brown=1, blue=9, red=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "pink": "The sampled ball is pink", "red": "The sampled ball is red", "orange": "The sampled ball is orange", "purple": "The sampled ball is purple", "brown": "The sampled ball is brown", "yellow": "The sampled ball is yellow", "black": "The sampled ball is black"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a brown ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: yellow, black, blue", "The drawn color is one of: orange, purple, red", "The drawn color is one of: pink, brown"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.23076923076923078, "pink": 0.05128205128205128, "red": 0.20512820512820512, "orange": 0.20512820512820512, "purple": 0.07692307692307693, "brown": 0.02564102564102564, "yellow": 0.02564102564102564, "black": 0.1794871794871795}, "focus": {"false": 0.9743589743589743, "true": 0.02564102564102564}, "group": {"0": 0.4358974358974359, "1": 0.48717948717948717, "2": 0.07692307692307693}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"7/39\"],[\"blue\",\"3/13\"],[\"brown\",\"1/39\"],[\"orange\",\"8/39\"],[\"pink\",\"2/39\"],[\"purple\",\"1/13\"],[\"red\",\"8/39\"],[\"yellow\",\"1/39\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0, "black": 0.01, "brown": 0.01, "purple": 0, "blue": 0.91, "yellow": 0.03, "red": 0.01, "orange": 0.03}, "focus": {"false": 0.95, "true": 0.05}, "group": {"0": 0.69, "1": 0.29, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:14", "state_id": "known_chance_v2:test:14", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: pink=7, gray=8, cyan=4, white=4, red=7, brown=9, blue=4, yellow=7, black=6, green=4, purple=8, orange=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"brown": "The sampled ball is brown", "orange": "The sampled ball is orange", "purple": "The sampled ball is purple", "black": "The sampled ball is black", "gray": "The sampled ball is gray", "red": "The sampled ball is red", "blue": "The sampled ball is blue", "yellow": "The sampled ball is yellow", "pink": "The sampled ball is pink", "white": "The sampled ball is white", "green": "The sampled ball is green", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a brown ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: pink, white, blue, green", "The drawn color is one of: gray, red, yellow, purple", "The drawn color is one of: cyan, brown, black, orange"]}}, "gold": {}, "gold_probs": {"outcome": {"brown": 0.1232876712328767, "orange": 0.0684931506849315, "purple": 0.1095890410958904, "black": 0.0821917808219178, "gray": 0.1095890410958904, "red": 0.0958904109589041, "blue": 0.0547945205479452, "yellow": 0.0958904109589041, "pink": 0.0958904109589041, "white": 0.0547945205479452, "green": 0.0547945205479452, "cyan": 0.0547945205479452}, "focus": {"false": 0.8767123287671232, "true": 0.1232876712328767}, "group": {"0": 0.2602739726027397, "1": 0.410958904109589, "2": 0.3287671232876712}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"6/73\"],[\"blue\",\"4/73\"],[\"brown\",\"9/73\"],[\"cyan\",\"4/73\"],[\"gray\",\"8/73\"],[\"green\",\"4/73\"],[\"orange\",\"5/73\"],[\"pink\",\"7/73\"],[\"purple\",\"8/73\"],[\"red\",\"7/73\"],[\"white\",\"4/73\"],[\"yellow\",\"7/73\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.01, "pink": 0.1, "green": 0, "orange": 0.01, "blue": 0.01, "cyan": 0.01, "red": 0.01, "white": 0.01, "yellow": 0.01, "brown": 0.77, "black": 0.01, "gray": 0.05}, "focus": {"false": 0.84, "true": 0.16}, "group": {"0": 0.02, "1": 0.67, "2": 0.31}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:15", "state_id": "known_chance_v2:test:15", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: orange=1, white=1, brown=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"orange": "The sampled ball is orange", "white": "The sampled ball is white", "brown": "The sampled ball is brown"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a white ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: orange", "The drawn color is one of: white", "The drawn color is one of: brown"]}}, "gold": {}, "gold_probs": {"outcome": {"orange": 0.1, "white": 0.1, "brown": 0.8}, "focus": {"false": 0.9, "true": 0.1}, "group": {"0": 0.1, "1": 0.1, "2": 0.8}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"4/5\"],[\"orange\",\"1/10\"],[\"white\",\"1/10\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"orange": 0.01, "brown": 0.99, "white": 0}, "focus": {"false": 0.91, "true": 0.09}, "group": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:16", "state_id": "known_chance_v2:test:16", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: black=7, white=4, brown=3, orange=9. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "brown": "The sampled ball is brown", "white": "The sampled ball is white", "orange": "The sampled ball is orange"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: black, orange", "The drawn color is one of: white", "The drawn color is one of: brown"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.30434782608695654, "brown": 0.13043478260869565, "white": 0.17391304347826086, "orange": 0.391304347826087}, "focus": {"false": 0.6956521739130435, "true": 0.30434782608695654}, "group": {"0": 0.6956521739130435, "1": 0.17391304347826086, "2": 0.13043478260869565}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"7/23\"],[\"brown\",\"3/23\"],[\"orange\",\"9/23\"],[\"white\",\"4/23\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0.01, "black": 0.24, "brown": 0.01, "orange": 0.74}, "focus": {"false": 0.74, "true": 0.26}, "group": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:17", "state_id": "known_chance_v2:test:17", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: blue=5, black=1, pink=7, green=5, red=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "blue": "The sampled ball is blue", "pink": "The sampled ball is pink", "green": "The sampled ball is green", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a red ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue, green", "The drawn color is one of: black, red", "The drawn color is one of: pink"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.041666666666666664, "blue": 0.20833333333333334, "pink": 0.2916666666666667, "green": 0.20833333333333334, "red": 0.25}, "focus": {"false": 0.75, "true": 0.25}, "group": {"0": 0.4166666666666667, "1": 0.2916666666666667, "2": 0.2916666666666667}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/24\"],[\"blue\",\"5/24\"],[\"green\",\"5/24\"],[\"pink\",\"7/24\"],[\"red\",\"1/4\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"black": 0.02, "blue": 0.12, "pink": 0.83, "green": 0.01, "red": 0.02}, "focus": {"false": 0.79, "true": 0.21}, "group": {"0": 0.3, "1": 0.02, "2": 0.68}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:19", "state_id": "known_chance_v2:test:19", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: black=9, brown=3, orange=9, blue=4, yellow=8, green=8, purple=8, gray=1, red=9, cyan=9, white=1, pink=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"pink": "The sampled ball is pink", "brown": "The sampled ball is brown", "blue": "The sampled ball is blue", "purple": "The sampled ball is purple", "orange": "The sampled ball is orange", "green": "The sampled ball is green", "yellow": "The sampled ball is yellow", "gray": "The sampled ball is gray", "white": "The sampled ball is white", "red": "The sampled ball is red", "black": "The sampled ball is black", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: black, blue, purple, cyan", "The drawn color is one of: brown, yellow, gray, white", "The drawn color is one of: orange, green, red, pink"]}}, "gold": {}, "gold_probs": {"outcome": {"pink": 0.08, "brown": 0.04, "blue": 0.05333333333333334, "purple": 0.10666666666666667, "orange": 0.12, "green": 0.10666666666666667, "yellow": 0.10666666666666667, "gray": 0.013333333333333334, "white": 0.013333333333333334, "red": 0.12, "black": 0.12, "cyan": 0.12}, "focus": {"false": 0.8933333333333333, "true": 0.10666666666666667}, "group": {"0": 0.4, "1": 0.17333333333333334, "2": 0.42666666666666664}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"3/25\"],[\"blue\",\"4/75\"],[\"brown\",\"1/25\"],[\"cyan\",\"3/25\"],[\"gray\",\"1/75\"],[\"green\",\"8/75\"],[\"orange\",\"3/25\"],[\"pink\",\"2/25\"],[\"purple\",\"8/75\"],[\"red\",\"3/25\"],[\"white\",\"1/75\"],[\"yellow\",\"8/75\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0.01, "orange": 0.05, "blue": 0.01, "yellow": 0.01, "red": 0.09, "gray": 0.02, "black": 0.74, "pink": 0.02, "cyan": 0.02, "purple": 0.01, "green": 0.01, "brown": 0.01}, "focus": {"false": 0.85, "true": 0.15}, "group": {"0": 0.43, "1": 0.03, "2": 0.54}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:18", "state_id": "known_chance_v2:test:18", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: brown=2, green=7, red=8, gray=8, orange=7, black=7, cyan=5, yellow=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"orange": "The sampled ball is orange", "gray": "The sampled ball is gray", "yellow": "The sampled ball is yellow", "red": "The sampled ball is red", "brown": "The sampled ball is brown", "green": "The sampled ball is green", "black": "The sampled ball is black", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: brown, gray, cyan", "The drawn color is one of: green, orange, yellow", "The drawn color is one of: red, black"]}}, "gold": {}, "gold_probs": {"outcome": {"orange": 0.14893617021276595, "gray": 0.1702127659574468, "yellow": 0.06382978723404255, "red": 0.1702127659574468, "brown": 0.0425531914893617, "green": 0.14893617021276595, "black": 0.14893617021276595, "cyan": 0.10638297872340426}, "focus": {"false": 0.8936170212765957, "true": 0.10638297872340426}, "group": {"0": 0.3191489361702128, "1": 0.36170212765957444, "2": 0.3191489361702128}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"7/47\"],[\"brown\",\"2/47\"],[\"cyan\",\"5/47\"],[\"gray\",\"8/47\"],[\"green\",\"7/47\"],[\"orange\",\"7/47\"],[\"red\",\"8/47\"],[\"yellow\",\"3/47\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"yellow": 0.01, "cyan": 0.01, "gray": 0.12, "black": 0, "green": 0.03, "red": 0.73, "brown": 0.09, "orange": 0.01}, "focus": {"false": 0.86, "true": 0.14}, "group": {"0": 0.03, "1": 0.07, "2": 0.9}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:20", "state_id": "known_chance_v2:test:20", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: blue=7, gray=9, red=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"red": "The sampled ball is red", "gray": "The sampled ball is gray", "blue": "The sampled ball is blue"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a gray ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue", "The drawn color is one of: gray", "The drawn color is one of: red"]}}, "gold": {}, "gold_probs": {"outcome": {"red": 0.1111111111111111, "gray": 0.5, "blue": 0.3888888888888889}, "focus": {"false": 0.5, "true": 0.5}, "group": {"0": 0.3888888888888889, "1": 0.5, "2": 0.1111111111111111}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"7/18\"],[\"gray\",\"1/2\"],[\"red\",\"1/9\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"gray": 0.72, "red": 0.02, "blue": 0.26}, "focus": {"false": 0.64, "true": 0.36}, "group": {"0": 0.02, "1": 0.97, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:21", "state_id": "known_chance_v2:test:21", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: pink=1, cyan=3, orange=7, white=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "pink": "The sampled ball is pink", "orange": "The sampled ball is orange", "white": "The sampled ball is white"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a white ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: pink, white", "The drawn color is one of: cyan", "The drawn color is one of: orange"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.15789473684210525, "pink": 0.05263157894736842, "orange": 0.3684210526315789, "white": 0.42105263157894735}, "focus": {"false": 0.5789473684210527, "true": 0.42105263157894735}, "group": {"0": 0.47368421052631576, "1": 0.15789473684210525, "2": 0.3684210526315789}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"3/19\"],[\"orange\",\"7/19\"],[\"pink\",\"1/19\"],[\"white\",\"8/19\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0.01, "white": 0.88, "orange": 0.03, "pink": 0.08}, "focus": {"false": 0.7, "true": 0.3}, "group": {"0": 0.45, "1": 0.01, "2": 0.54}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:23", "state_id": "known_chance_v2:test:23", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: red=7, blue=8, purple=1, gray=5, orange=6, black=9, cyan=1, brown=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"brown": "The sampled ball is brown", "gray": "The sampled ball is gray", "blue": "The sampled ball is blue", "orange": "The sampled ball is orange", "red": "The sampled ball is red", "purple": "The sampled ball is purple", "cyan": "The sampled ball is cyan", "black": "The sampled ball is black"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a gray ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: red, gray, cyan", "The drawn color is one of: blue, orange, brown", "The drawn color is one of: purple, black"]}}, "gold": {}, "gold_probs": {"outcome": {"brown": 0.13953488372093023, "gray": 0.11627906976744186, "blue": 0.18604651162790697, "orange": 0.13953488372093023, "red": 0.16279069767441862, "purple": 0.023255813953488372, "cyan": 0.023255813953488372, "black": 0.20930232558139536}, "focus": {"false": 0.8837209302325582, "true": 0.11627906976744186}, "group": {"0": 0.3023255813953488, "1": 0.46511627906976744, "2": 0.23255813953488375}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"9/43\"],[\"blue\",\"8/43\"],[\"brown\",\"6/43\"],[\"cyan\",\"1/43\"],[\"gray\",\"5/43\"],[\"orange\",\"6/43\"],[\"purple\",\"1/43\"],[\"red\",\"7/43\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.01, "cyan": 0, "brown": 0, "blue": 0.04, "gray": 0, "red": 0.04, "orange": 0, "black": 0.91}, "focus": {"false": 0.84, "true": 0.16}, "group": {"0": 0.03, "1": 0.8, "2": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:22", "state_id": "known_chance_v2:test:22", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: black=7, yellow=4, green=8, gray=9, cyan=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "black": "The sampled ball is black", "cyan": "The sampled ball is cyan", "gray": "The sampled ball is gray", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: black, gray", "The drawn color is one of: yellow, cyan", "The drawn color is one of: green"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.22857142857142856, "black": 0.2, "cyan": 0.2, "gray": 0.2571428571428571, "yellow": 0.11428571428571428}, "focus": {"false": 0.8, "true": 0.2}, "group": {"0": 0.45714285714285713, "1": 0.3142857142857143, "2": 0.22857142857142856}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/5\"],[\"cyan\",\"1/5\"],[\"gray\",\"9/35\"],[\"green\",\"8/35\"],[\"yellow\",\"4/35\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0.02, "gray": 0.92, "cyan": 0, "yellow": 0.01, "black": 0.05}, "focus": {"false": 0.81, "true": 0.19}, "group": {"0": 0.92, "1": 0.01, "2": 0.07}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:24", "state_id": "known_chance_v2:test:24", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: white=6, purple=6, red=8, black=3, pink=6, green=1, blue=7, yellow=3, brown=4, cyan=3, orange=8, gray=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "brown": "The sampled ball is brown", "yellow": "The sampled ball is yellow", "orange": "The sampled ball is orange", "red": "The sampled ball is red", "black": "The sampled ball is black", "blue": "The sampled ball is blue", "gray": "The sampled ball is gray", "purple": "The sampled ball is purple", "pink": "The sampled ball is pink", "white": "The sampled ball is white", "green": "The sampled ball is green"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a orange ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white, black, blue, cyan", "The drawn color is one of: purple, pink, yellow, orange", "The drawn color is one of: red, green, brown, gray"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.05263157894736842, "brown": 0.07017543859649122, "yellow": 0.05263157894736842, "orange": 0.14035087719298245, "red": 0.14035087719298245, "black": 0.05263157894736842, "blue": 0.12280701754385964, "gray": 0.03508771929824561, "purple": 0.10526315789473684, "pink": 0.10526315789473684, "white": 0.10526315789473684, "green": 0.017543859649122806}, "focus": {"false": 0.8596491228070176, "true": 0.14035087719298245}, "group": {"0": 0.3333333333333333, "1": 0.40350877192982454, "2": 0.2631578947368421}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/19\"],[\"blue\",\"7/57\"],[\"brown\",\"4/57\"],[\"cyan\",\"1/19\"],[\"gray\",\"2/57\"],[\"green\",\"1/57\"],[\"orange\",\"8/57\"],[\"pink\",\"2/19\"],[\"purple\",\"2/19\"],[\"red\",\"8/57\"],[\"white\",\"2/19\"],[\"yellow\",\"1/19\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.01, "green": 0.1, "gray": 0.01, "cyan": 0.01, "blue": 0.04, "white": 0.38, "yellow": 0, "orange": 0.04, "purple": 0.01, "red": 0.38, "brown": 0.01, "black": 0.01}, "focus": {"false": 0.86, "true": 0.14}, "group": {"0": 0.36, "1": 0.44, "2": 0.2}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:25", "state_id": "known_chance_v2:test:25", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: green=6, red=1, orange=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"orange": "The sampled ball is orange", "red": "The sampled ball is red", "green": "The sampled ball is green"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a green ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: green", "The drawn color is one of: red", "The drawn color is one of: orange"]}}, "gold": {}, "gold_probs": {"outcome": {"orange": 0.5333333333333333, "red": 0.06666666666666667, "green": 0.4}, "focus": {"false": 0.6, "true": 0.4}, "group": {"0": 0.4, "1": 0.06666666666666667, "2": 0.5333333333333333}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"green\",\"2/5\"],[\"orange\",\"8/15\"],[\"red\",\"1/15\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0.21, "orange": 0.77, "red": 0.02}, "focus": {"false": 0.6799999999999999, "true": 0.32}, "group": {"0": 0.08, "1": 0.01, "2": 0.91}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:26", "state_id": "known_chance_v2:test:26", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: white=5, gray=6, blue=5, purple=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "gray": "The sampled ball is gray", "white": "The sampled ball is white", "purple": "The sampled ball is purple"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white, purple", "The drawn color is one of: gray", "The drawn color is one of: blue"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.21739130434782608, "gray": 0.2608695652173913, "white": 0.21739130434782608, "purple": 0.30434782608695654}, "focus": {"false": 0.782608695652174, "true": 0.21739130434782608}, "group": {"0": 0.5217391304347826, "1": 0.2608695652173913, "2": 0.21739130434782608}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"5/23\"],[\"gray\",\"6/23\"],[\"purple\",\"7/23\"],[\"white\",\"5/23\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0.2, "blue": 0.02, "purple": 0.53, "gray": 0.25}, "focus": {"false": 0.8, "true": 0.2}, "group": {"0": 0.76, "1": 0.23, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:28", "state_id": "known_chance_v2:test:28", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: pink=4, gray=7, white=7, purple=4, cyan=3, green=8, orange=4, brown=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "green": "The sampled ball is green", "gray": "The sampled ball is gray", "orange": "The sampled ball is orange", "white": "The sampled ball is white", "pink": "The sampled ball is pink", "brown": "The sampled ball is brown", "purple": "The sampled ball is purple"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a gray ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: pink, purple, orange", "The drawn color is one of: gray, cyan, brown", "The drawn color is one of: white, green"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.07142857142857142, "green": 0.19047619047619047, "gray": 0.16666666666666666, "orange": 0.09523809523809523, "white": 0.16666666666666666, "pink": 0.09523809523809523, "brown": 0.11904761904761904, "purple": 0.09523809523809523}, "focus": {"false": 0.8333333333333334, "true": 0.16666666666666666}, "group": {"0": 0.2857142857142857, "1": 0.3571428571428571, "2": 0.3571428571428571}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"5/42\"],[\"cyan\",\"1/14\"],[\"gray\",\"1/6\"],[\"green\",\"4/21\"],[\"orange\",\"2/21\"],[\"pink\",\"2/21\"],[\"purple\",\"2/21\"],[\"white\",\"1/6\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"brown": 0.01, "purple": 0, "orange": 0.01, "pink": 0.02, "gray": 0.07, "cyan": 0.02, "green": 0.85, "white": 0.02}, "focus": {"false": 0.8, "true": 0.2}, "group": {"0": 0.08, "1": 0.15, "2": 0.77}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:27", "state_id": "known_chance_v2:test:27", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: brown=9, blue=8, orange=3, purple=9, black=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"purple": "The sampled ball is purple", "orange": "The sampled ball is orange", "black": "The sampled ball is black", "brown": "The sampled ball is brown", "blue": "The sampled ball is blue"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a orange ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: brown, purple", "The drawn color is one of: blue, black", "The drawn color is one of: orange"]}}, "gold": {}, "gold_probs": {"outcome": {"purple": 0.2571428571428571, "orange": 0.08571428571428572, "black": 0.17142857142857143, "brown": 0.2571428571428571, "blue": 0.22857142857142856}, "focus": {"false": 0.9142857142857143, "true": 0.08571428571428572}, "group": {"0": 0.5142857142857142, "1": 0.4, "2": 0.08571428571428572}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"6/35\"],[\"blue\",\"8/35\"],[\"brown\",\"9/35\"],[\"orange\",\"3/35\"],[\"purple\",\"9/35\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"blue": 0.02, "black": 0.02, "brown": 0.82, "purple": 0.11, "orange": 0.03}, "focus": {"false": 0.88, "true": 0.12}, "group": {"0": 0.98, "1": 0.01, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:31", "state_id": "known_chance_v2:test:31", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: blue=6, gray=8, brown=8, orange=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "brown": "The sampled ball is brown", "orange": "The sampled ball is orange", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a orange ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue, orange", "The drawn color is one of: gray", "The drawn color is one of: brown"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.21428571428571427, "brown": 0.2857142857142857, "orange": 0.21428571428571427, "gray": 0.2857142857142857}, "focus": {"false": 0.7857142857142857, "true": 0.21428571428571427}, "group": {"0": 0.42857142857142855, "1": 0.2857142857142857, "2": 0.2857142857142857}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"3/14\"],[\"brown\",\"2/7\"],[\"gray\",\"2/7\"],[\"orange\",\"3/14\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"gray": 0.4, "orange": 0.02, "blue": 0.51, "brown": 0.07}, "focus": {"false": 0.81, "true": 0.19}, "group": {"0": 0.26, "1": 0.65, "2": 0.09}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:29", "state_id": "known_chance_v2:test:29", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: black=6, green=8, red=5, orange=5, cyan=5, white=7, blue=4, gray=9, yellow=1, brown=5, pink=5, purple=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"red": "The sampled ball is red", "blue": "The sampled ball is blue", "white": "The sampled ball is white", "yellow": "The sampled ball is yellow", "black": "The sampled ball is black", "purple": "The sampled ball is purple", "brown": "The sampled ball is brown", "orange": "The sampled ball is orange", "gray": "The sampled ball is gray", "cyan": "The sampled ball is cyan", "pink": "The sampled ball is pink", "green": "The sampled ball is green"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: black, orange, blue, brown", "The drawn color is one of: green, cyan, gray, pink", "The drawn color is one of: red, white, yellow, purple"]}}, "gold": {}, "gold_probs": {"outcome": {"red": 0.08196721311475409, "blue": 0.06557377049180328, "white": 0.11475409836065574, "yellow": 0.01639344262295082, "black": 0.09836065573770492, "purple": 0.01639344262295082, "brown": 0.08196721311475409, "orange": 0.08196721311475409, "gray": 0.14754098360655737, "cyan": 0.08196721311475409, "pink": 0.08196721311475409, "green": 0.13114754098360656}, "focus": {"false": 0.9016393442622951, "true": 0.09836065573770492}, "group": {"0": 0.32786885245901637, "1": 0.4426229508196721, "2": 0.22950819672131148}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"6/61\"],[\"blue\",\"4/61\"],[\"brown\",\"5/61\"],[\"cyan\",\"5/61\"],[\"gray\",\"9/61\"],[\"green\",\"8/61\"],[\"orange\",\"5/61\"],[\"pink\",\"5/61\"],[\"purple\",\"1/61\"],[\"red\",\"5/61\"],[\"white\",\"7/61\"],[\"yellow\",\"1/61\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0, "cyan": 0, "yellow": 0.01, "white": 0.01, "gray": 0.84, "purple": 0.01, "black": 0.06, "green": 0.07, "blue": 0, "red": 0, "brown": 0, "orange": 0}, "focus": {"false": 0.85, "true": 0.15}, "group": {"0": 0.08, "1": 0.88, "2": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:30", "state_id": "known_chance_v2:test:30", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: purple=9, gray=3, red=9. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"gray": "The sampled ball is gray", "purple": "The sampled ball is purple", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a gray ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: purple", "The drawn color is one of: gray", "The drawn color is one of: red"]}}, "gold": {}, "gold_probs": {"outcome": {"gray": 0.14285714285714285, "purple": 0.42857142857142855, "red": 0.42857142857142855}, "focus": {"false": 0.8571428571428572, "true": 0.14285714285714285}, "group": {"0": 0.42857142857142855, "1": 0.14285714285714285, "2": 0.42857142857142855}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"gray\",\"1/7\"],[\"purple\",\"3/7\"],[\"red\",\"3/7\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.89, "red": 0.06, "gray": 0.05}, "focus": {"false": 0.81, "true": 0.19}, "group": {"0": 0.82, "1": 0.04, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:32", "state_id": "known_chance_v2:test:32", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: blue=6, cyan=4, red=1, black=4, green=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"red": "The sampled ball is red", "green": "The sampled ball is green", "blue": "The sampled ball is blue", "cyan": "The sampled ball is cyan", "black": "The sampled ball is black"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue, black", "The drawn color is one of: cyan, green", "The drawn color is one of: red"]}}, "gold": {}, "gold_probs": {"outcome": {"red": 0.05555555555555555, "green": 0.16666666666666666, "blue": 0.3333333333333333, "cyan": 0.2222222222222222, "black": 0.2222222222222222}, "focus": {"false": 0.6666666666666667, "true": 0.3333333333333333}, "group": {"0": 0.5555555555555556, "1": 0.38888888888888884, "2": 0.05555555555555555}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"2/9\"],[\"blue\",\"1/3\"],[\"cyan\",\"2/9\"],[\"green\",\"1/6\"],[\"red\",\"1/18\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0.01, "black": 0, "blue": 0.98, "green": 0, "red": 0.01}, "focus": {"false": 0.7, "true": 0.3}, "group": {"0": 0.97, "1": 0.02, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:33", "state_id": "known_chance_v2:test:33", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: blue=5, gray=8, brown=6, pink=7, black=2, green=8, cyan=8, red=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "green": "The sampled ball is green", "blue": "The sampled ball is blue", "black": "The sampled ball is black", "pink": "The sampled ball is pink", "brown": "The sampled ball is brown", "red": "The sampled ball is red", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue, pink, cyan", "The drawn color is one of: gray, black, red", "The drawn color is one of: brown, green"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.15384615384615385, "green": 0.15384615384615385, "blue": 0.09615384615384616, "black": 0.038461538461538464, "pink": 0.1346153846153846, "brown": 0.11538461538461539, "red": 0.15384615384615385, "gray": 0.15384615384615385}, "focus": {"false": 0.9615384615384616, "true": 0.038461538461538464}, "group": {"0": 0.38461538461538464, "1": 0.34615384615384615, "2": 0.2692307692307693}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/26\"],[\"blue\",\"5/52\"],[\"brown\",\"3/26\"],[\"cyan\",\"2/13\"],[\"gray\",\"2/13\"],[\"green\",\"2/13\"],[\"pink\",\"7/52\"],[\"red\",\"2/13\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0.04, "pink": 0.02, "gray": 0.35, "red": 0.05, "cyan": 0.03, "brown": 0.02, "blue": 0.38, "black": 0.11}, "focus": {"false": 0.89, "true": 0.11}, "group": {"0": 0.12, "1": 0.7, "2": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:34", "state_id": "known_chance_v2:test:34", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: blue=1, yellow=7, black=3, gray=1, orange=7, purple=1, red=1, brown=9, white=1, pink=6, cyan=4, green=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "green": "The sampled ball is green", "gray": "The sampled ball is gray", "brown": "The sampled ball is brown", "red": "The sampled ball is red", "yellow": "The sampled ball is yellow", "purple": "The sampled ball is purple", "white": "The sampled ball is white", "pink": "The sampled ball is pink", "orange": "The sampled ball is orange", "blue": "The sampled ball is blue", "black": "The sampled ball is black"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue, gray, red, pink", "The drawn color is one of: yellow, orange, brown, cyan", "The drawn color is one of: black, purple, white, green"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.08695652173913043, "green": 0.10869565217391304, "gray": 0.021739130434782608, "brown": 0.1956521739130435, "red": 0.021739130434782608, "yellow": 0.15217391304347827, "purple": 0.021739130434782608, "white": 0.021739130434782608, "pink": 0.13043478260869565, "orange": 0.15217391304347827, "blue": 0.021739130434782608, "black": 0.06521739130434782}, "focus": {"false": 0.8478260869565217, "true": 0.15217391304347827}, "group": {"0": 0.19565217391304346, "1": 0.5869565217391305, "2": 0.21739130434782608}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"3/46\"],[\"blue\",\"1/46\"],[\"brown\",\"9/46\"],[\"cyan\",\"2/23\"],[\"gray\",\"1/46\"],[\"green\",\"5/46\"],[\"orange\",\"7/46\"],[\"pink\",\"3/23\"],[\"purple\",\"1/46\"],[\"red\",\"1/46\"],[\"white\",\"1/46\"],[\"yellow\",\"7/46\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"blue": 0.01, "red": 0, "gray": 0, "orange": 0.01, "green": 0, "white": 0, "black": 0, "yellow": 0.02, "cyan": 0, "purple": 0, "brown": 0.96, "pink": 0}, "focus": {"false": 0.81, "true": 0.19}, "group": {"0": 0, "1": 1, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:36", "state_id": "known_chance_v2:test:36", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: pink=5, white=9, blue=6, red=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "pink": "The sampled ball is pink", "white": "The sampled ball is white", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: pink, red", "The drawn color is one of: white", "The drawn color is one of: blue"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.2727272727272727, "pink": 0.22727272727272727, "white": 0.4090909090909091, "red": 0.09090909090909091}, "focus": {"false": 0.7272727272727273, "true": 0.2727272727272727}, "group": {"0": 0.3181818181818182, "1": 0.4090909090909091, "2": 0.2727272727272727}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"3/11\"],[\"pink\",\"5/22\"],[\"red\",\"1/11\"],[\"white\",\"9/22\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"red": 0.01, "white": 0.89, "pink": 0.09, "blue": 0.01}, "focus": {"false": 0.79, "true": 0.21}, "group": {"0": 0.03, "1": 0.88, "2": 0.09}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:37", "state_id": "known_chance_v2:test:37", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: green=8, black=9, yellow=9, purple=6, orange=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"yellow": "The sampled ball is yellow", "green": "The sampled ball is green", "orange": "The sampled ball is orange", "purple": "The sampled ball is purple", "black": "The sampled ball is black"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a orange ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: green, purple", "The drawn color is one of: black, orange", "The drawn color is one of: yellow"]}}, "gold": {}, "gold_probs": {"outcome": {"yellow": 0.25, "green": 0.2222222222222222, "orange": 0.1111111111111111, "purple": 0.16666666666666666, "black": 0.25}, "focus": {"false": 0.8888888888888888, "true": 0.1111111111111111}, "group": {"0": 0.38888888888888884, "1": 0.3611111111111111, "2": 0.25}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/4\"],[\"green\",\"2/9\"],[\"orange\",\"1/9\"],[\"purple\",\"1/6\"],[\"yellow\",\"1/4\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"black": 0.7, "green": 0.21, "orange": 0.02, "purple": 0.01, "yellow": 0.06}, "focus": {"false": 0.87, "true": 0.13}, "group": {"0": 0.25, "1": 0.53, "2": 0.22}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:35", "state_id": "known_chance_v2:test:35", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: red=9, purple=4, blue=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"purple": "The sampled ball is purple", "blue": "The sampled ball is blue", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a purple ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: red", "The drawn color is one of: purple", "The drawn color is one of: blue"]}}, "gold": {}, "gold_probs": {"outcome": {"purple": 0.2857142857142857, "blue": 0.07142857142857142, "red": 0.6428571428571429}, "focus": {"false": 0.7142857142857143, "true": 0.2857142857142857}, "group": {"0": 0.6428571428571429, "1": 0.2857142857142857, "2": 0.07142857142857142}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"1/14\"],[\"purple\",\"2/7\"],[\"red\",\"9/14\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.01, "red": 0.98, "blue": 0.01}, "focus": {"false": 0.75, "true": 0.25}, "group": {"0": 0.98, "1": 0.01, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:38", "state_id": "known_chance_v2:test:38", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: green=3, brown=1, pink=2, orange=7, yellow=5, gray=9, cyan=5, white=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "green": "The sampled ball is green", "gray": "The sampled ball is gray", "yellow": "The sampled ball is yellow", "brown": "The sampled ball is brown", "white": "The sampled ball is white", "pink": "The sampled ball is pink", "orange": "The sampled ball is orange"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a white ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: green, orange, cyan", "The drawn color is one of: brown, yellow, white", "The drawn color is one of: pink, gray"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.14705882352941177, "green": 0.08823529411764706, "gray": 0.2647058823529412, "yellow": 0.14705882352941177, "brown": 0.029411764705882353, "white": 0.058823529411764705, "pink": 0.058823529411764705, "orange": 0.20588235294117646}, "focus": {"false": 0.9411764705882353, "true": 0.058823529411764705}, "group": {"0": 0.4411764705882353, "1": 0.23529411764705882, "2": 0.3235294117647059}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"1/34\"],[\"cyan\",\"5/34\"],[\"gray\",\"9/34\"],[\"green\",\"3/34\"],[\"orange\",\"7/34\"],[\"pink\",\"1/17\"],[\"white\",\"1/17\"],[\"yellow\",\"5/34\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0, "orange": 0.07, "green": 0.03, "gray": 0.88, "yellow": 0, "cyan": 0, "pink": 0, "brown": 0.02}, "focus": {"false": 0.89, "true": 0.11}, "group": {"0": 0.37, "1": 0.01, "2": 0.62}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:40", "state_id": "known_chance_v2:test:40", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: cyan=2, blue=2, white=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"white": "The sampled ball is white", "blue": "The sampled ball is blue", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: cyan", "The drawn color is one of: blue", "The drawn color is one of: white"]}}, "gold": {}, "gold_probs": {"outcome": {"white": 0.6363636363636364, "blue": 0.18181818181818182, "cyan": 0.18181818181818182}, "focus": {"false": 0.8181818181818181, "true": 0.18181818181818182}, "group": {"0": 0.18181818181818182, "1": 0.18181818181818182, "2": 0.6363636363636364}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"2/11\"],[\"cyan\",\"2/11\"],[\"white\",\"7/11\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0.01, "blue": 0, "white": 0.99}, "focus": {"false": 0.79, "true": 0.21}, "group": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:39", "state_id": "known_chance_v2:test:39", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: red=6, yellow=8, cyan=3, pink=4, brown=2, purple=1, green=4, black=9, orange=7, blue=5, white=1, gray=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"yellow": "The sampled ball is yellow", "white": "The sampled ball is white", "gray": "The sampled ball is gray", "blue": "The sampled ball is blue", "purple": "The sampled ball is purple", "black": "The sampled ball is black", "red": "The sampled ball is red", "green": "The sampled ball is green", "pink": "The sampled ball is pink", "orange": "The sampled ball is orange", "cyan": "The sampled ball is cyan", "brown": "The sampled ball is brown"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a orange ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: red, pink, green, blue", "The drawn color is one of: yellow, brown, black, white", "The drawn color is one of: cyan, purple, orange, gray"]}}, "gold": {}, "gold_probs": {"outcome": {"yellow": 0.14814814814814814, "white": 0.018518518518518517, "gray": 0.07407407407407407, "blue": 0.09259259259259259, "purple": 0.018518518518518517, "black": 0.16666666666666666, "red": 0.1111111111111111, "green": 0.07407407407407407, "pink": 0.07407407407407407, "orange": 0.12962962962962962, "cyan": 0.05555555555555555, "brown": 0.037037037037037035}, "focus": {"false": 0.8703703703703703, "true": 0.12962962962962962}, "group": {"0": 0.35185185185185186, "1": 0.37037037037037035, "2": 0.2777777777777778}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/6\"],[\"blue\",\"5/54\"],[\"brown\",\"1/27\"],[\"cyan\",\"1/18\"],[\"gray\",\"2/27\"],[\"green\",\"2/27\"],[\"orange\",\"7/54\"],[\"pink\",\"2/27\"],[\"purple\",\"1/54\"],[\"red\",\"1/9\"],[\"white\",\"1/54\"],[\"yellow\",\"4/27\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0, "green": 0, "blue": 0, "purple": 0, "pink": 0, "white": 0, "red": 0.05, "brown": 0, "black": 0.93, "orange": 0, "gray": 0, "yellow": 0.01}, "focus": {"false": 0.85, "true": 0.15}, "group": {"0": 0.12, "1": 0.85, "2": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:43", "state_id": "known_chance_v2:test:43", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: green=7, cyan=8, blue=3, yellow=4, red=4, white=4, gray=6, pink=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"red": "The sampled ball is red", "white": "The sampled ball is white", "gray": "The sampled ball is gray", "yellow": "The sampled ball is yellow", "cyan": "The sampled ball is cyan", "green": "The sampled ball is green", "pink": "The sampled ball is pink", "blue": "The sampled ball is blue"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a gray ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: green, yellow, gray", "The drawn color is one of: cyan, red, pink", "The drawn color is one of: blue, white"]}}, "gold": {}, "gold_probs": {"outcome": {"red": 0.09302325581395349, "white": 0.09302325581395349, "gray": 0.13953488372093023, "yellow": 0.09302325581395349, "cyan": 0.18604651162790697, "green": 0.16279069767441862, "pink": 0.16279069767441862, "blue": 0.06976744186046512}, "focus": {"false": 0.8604651162790697, "true": 0.13953488372093023}, "group": {"0": 0.39534883720930236, "1": 0.4418604651162791, "2": 0.16279069767441862}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"3/43\"],[\"cyan\",\"8/43\"],[\"gray\",\"6/43\"],[\"green\",\"7/43\"],[\"pink\",\"7/43\"],[\"red\",\"4/43\"],[\"white\",\"4/43\"],[\"yellow\",\"4/43\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"blue": 0.01, "red": 0, "white": 0, "gray": 0, "green": 0.06, "yellow": 0, "pink": 0, "cyan": 0.93}, "focus": {"false": 0.83, "true": 0.17}, "group": {"0": 0.5, "1": 0.48, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:41", "state_id": "known_chance_v2:test:41", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: yellow=7, red=5, green=4, white=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "white": "The sampled ball is white", "yellow": "The sampled ball is yellow", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: yellow, white", "The drawn color is one of: red", "The drawn color is one of: green"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.16666666666666666, "white": 0.3333333333333333, "yellow": 0.2916666666666667, "red": 0.20833333333333334}, "focus": {"false": 0.7083333333333333, "true": 0.2916666666666667}, "group": {"0": 0.625, "1": 0.20833333333333334, "2": 0.16666666666666666}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"green\",\"1/6\"],[\"red\",\"5/24\"],[\"white\",\"1/3\"],[\"yellow\",\"7/24\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0.79, "yellow": 0.2, "red": 0, "green": 0.01}, "focus": {"false": 0.73, "true": 0.27}, "group": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:46", "state_id": "known_chance_v2:test:46", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: blue=3, gray=8, brown=3, orange=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"brown": "The sampled ball is brown", "blue": "The sampled ball is blue", "gray": "The sampled ball is gray", "orange": "The sampled ball is orange"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a brown ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue, orange", "The drawn color is one of: gray", "The drawn color is one of: brown"]}}, "gold": {}, "gold_probs": {"outcome": {"brown": 0.17647058823529413, "blue": 0.17647058823529413, "gray": 0.47058823529411764, "orange": 0.17647058823529413}, "focus": {"false": 0.8235294117647058, "true": 0.17647058823529413}, "group": {"0": 0.35294117647058826, "1": 0.47058823529411764, "2": 0.17647058823529413}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"3/17\"],[\"brown\",\"3/17\"],[\"gray\",\"8/17\"],[\"orange\",\"3/17\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"gray": 0.95, "blue": 0.05, "brown": 0, "orange": 0}, "focus": {"false": 0.81, "true": 0.19}, "group": {"0": 0.03, "1": 0.96, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:45", "state_id": "known_chance_v2:test:45", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: purple=7, pink=1, green=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "purple": "The sampled ball is purple", "pink": "The sampled ball is pink"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a purple ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: purple", "The drawn color is one of: pink", "The drawn color is one of: green"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.2727272727272727, "purple": 0.6363636363636364, "pink": 0.09090909090909091}, "focus": {"false": 0.36363636363636365, "true": 0.6363636363636364}, "group": {"0": 0.6363636363636364, "1": 0.09090909090909091, "2": 0.2727272727272727}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"green\",\"3/11\"],[\"pink\",\"1/11\"],[\"purple\",\"7/11\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0, "green": 0.01, "purple": 0.99}, "focus": {"false": 0.45999999999999996, "true": 0.54}, "group": {"0": 0.99, "1": 0, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:44", "state_id": "known_chance_v2:test:44", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: cyan=7, yellow=3, white=7, red=5, gray=8, green=1, blue=6, pink=1, orange=4, black=1, purple=5, brown=9. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"orange": "The sampled ball is orange", "yellow": "The sampled ball is yellow", "purple": "The sampled ball is purple", "red": "The sampled ball is red", "blue": "The sampled ball is blue", "gray": "The sampled ball is gray", "white": "The sampled ball is white", "pink": "The sampled ball is pink", "black": "The sampled ball is black", "cyan": "The sampled ball is cyan", "brown": "The sampled ball is brown", "green": "The sampled ball is green"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a gray ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: cyan, red, blue, black", "The drawn color is one of: yellow, gray, pink, purple", "The drawn color is one of: white, green, orange, brown"]}}, "gold": {}, "gold_probs": {"outcome": {"orange": 0.07017543859649122, "yellow": 0.05263157894736842, "purple": 0.08771929824561403, "red": 0.08771929824561403, "blue": 0.10526315789473684, "gray": 0.14035087719298245, "white": 0.12280701754385964, "pink": 0.017543859649122806, "black": 0.017543859649122806, "cyan": 0.12280701754385964, "brown": 0.15789473684210525, "green": 0.017543859649122806}, "focus": {"false": 0.8596491228070176, "true": 0.14035087719298245}, "group": {"0": 0.3333333333333333, "1": 0.2982456140350877, "2": 0.3684210526315789}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/57\"],[\"blue\",\"2/19\"],[\"brown\",\"3/19\"],[\"cyan\",\"7/57\"],[\"gray\",\"8/57\"],[\"green\",\"1/57\"],[\"orange\",\"4/57\"],[\"pink\",\"1/57\"],[\"purple\",\"5/57\"],[\"red\",\"5/57\"],[\"white\",\"7/57\"],[\"yellow\",\"1/19\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0, "yellow": 0, "cyan": 0.05, "white": 0.01, "purple": 0, "gray": 0.02, "black": 0, "blue": 0, "green": 0, "red": 0, "brown": 0.92, "orange": 0}, "focus": {"false": 0.84, "true": 0.16}, "group": {"0": 0.08, "1": 0.15, "2": 0.77}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:47", "state_id": "known_chance_v2:test:47", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: green=1, brown=7, orange=6, white=3, pink=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"orange": "The sampled ball is orange", "green": "The sampled ball is green", "white": "The sampled ball is white", "pink": "The sampled ball is pink", "brown": "The sampled ball is brown"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a pink ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: green, white", "The drawn color is one of: brown, pink", "The drawn color is one of: orange"]}}, "gold": {}, "gold_probs": {"outcome": {"orange": 0.25, "green": 0.041666666666666664, "white": 0.125, "pink": 0.2916666666666667, "brown": 0.2916666666666667}, "focus": {"false": 0.7083333333333333, "true": 0.2916666666666667}, "group": {"0": 0.16666666666666666, "1": 0.5833333333333334, "2": 0.25}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"7/24\"],[\"green\",\"1/24\"],[\"orange\",\"1/4\"],[\"pink\",\"7/24\"],[\"white\",\"1/8\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0.01, "orange": 0.02, "pink": 0.05, "green": 0.2, "brown": 0.72}, "focus": {"false": 0.75, "true": 0.25}, "group": {"0": 0.01, "1": 0.99, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:49", "state_id": "known_chance_v2:test:49", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: red=5, gray=1, black=6, cyan=4, white=6, blue=1, brown=3, pink=4, purple=9, yellow=6, orange=8, green=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"purple": "The sampled ball is purple", "cyan": "The sampled ball is cyan", "yellow": "The sampled ball is yellow", "brown": "The sampled ball is brown", "white": "The sampled ball is white", "gray": "The sampled ball is gray", "orange": "The sampled ball is orange", "green": "The sampled ball is green", "black": "The sampled ball is black", "pink": "The sampled ball is pink", "blue": "The sampled ball is blue", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: red, cyan, brown, yellow", "The drawn color is one of: gray, white, pink, orange", "The drawn color is one of: black, blue, purple, green"]}}, "gold": {}, "gold_probs": {"outcome": {"purple": 0.15789473684210525, "cyan": 0.07017543859649122, "yellow": 0.10526315789473684, "brown": 0.05263157894736842, "white": 0.10526315789473684, "gray": 0.017543859649122806, "orange": 0.14035087719298245, "green": 0.07017543859649122, "black": 0.10526315789473684, "pink": 0.07017543859649122, "blue": 0.017543859649122806, "red": 0.08771929824561403}, "focus": {"false": 0.9298245614035088, "true": 0.07017543859649122}, "group": {"0": 0.3157894736842105, "1": 0.3333333333333333, "2": 0.3508771929824561}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"2/19\"],[\"blue\",\"1/57\"],[\"brown\",\"1/19\"],[\"cyan\",\"4/57\"],[\"gray\",\"1/57\"],[\"green\",\"4/57\"],[\"orange\",\"8/57\"],[\"pink\",\"4/57\"],[\"purple\",\"3/19\"],[\"red\",\"5/57\"],[\"white\",\"2/19\"],[\"yellow\",\"2/19\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.83, "cyan": 0.01, "white": 0.01, "pink": 0, "gray": 0.01, "brown": 0, "blue": 0.01, "yellow": 0.01, "red": 0.07, "orange": 0.02, "green": 0, "black": 0.03}, "focus": {"false": 0.88, "true": 0.12}, "group": {"0": 0.04, "1": 0.11, "2": 0.85}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:50", "state_id": "known_chance_v2:test:50", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: brown=7, red=5, yellow=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"red": "The sampled ball is red", "brown": "The sampled ball is brown", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a brown ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: brown", "The drawn color is one of: red", "The drawn color is one of: yellow"]}}, "gold": {}, "gold_probs": {"outcome": {"red": 0.2777777777777778, "brown": 0.3888888888888889, "yellow": 0.3333333333333333}, "focus": {"false": 0.6111111111111112, "true": 0.3888888888888889}, "group": {"0": 0.3888888888888889, "1": 0.2777777777777778, "2": 0.3333333333333333}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"7/18\"],[\"red\",\"5/18\"],[\"yellow\",\"1/3\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"red": 0.01, "brown": 0.98, "yellow": 0.01}, "focus": {"false": 0.64, "true": 0.36}, "group": {"0": 0.93, "1": 0.02, "2": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:42", "state_id": "known_chance_v2:test:42", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: gray=6, yellow=5, pink=8, orange=5, white=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"pink": "The sampled ball is pink", "yellow": "The sampled ball is yellow", "white": "The sampled ball is white", "orange": "The sampled ball is orange", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a white ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: gray, orange", "The drawn color is one of: yellow, white", "The drawn color is one of: pink"]}}, "gold": {}, "gold_probs": {"outcome": {"pink": 0.3076923076923077, "yellow": 0.19230769230769232, "white": 0.07692307692307693, "orange": 0.19230769230769232, "gray": 0.23076923076923078}, "focus": {"false": 0.9230769230769231, "true": 0.07692307692307693}, "group": {"0": 0.42307692307692313, "1": 0.2692307692307693, "2": 0.3076923076923077}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"gray\",\"3/13\"],[\"orange\",\"5/26\"],[\"pink\",\"4/13\"],[\"white\",\"1/13\"],[\"yellow\",\"5/26\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"gray": 0.13, "orange": 0, "white": 0.02, "pink": 0.84, "yellow": 0.01}, "focus": {"false": 0.89, "true": 0.11}, "group": {"0": 0.3, "1": 0.02, "2": 0.68}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:51", "state_id": "known_chance_v2:test:51", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: brown=3, pink=9, black=3, purple=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "purple": "The sampled ball is purple", "pink": "The sampled ball is pink", "brown": "The sampled ball is brown"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a pink ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: brown, purple", "The drawn color is one of: pink", "The drawn color is one of: black"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.17647058823529413, "purple": 0.11764705882352941, "pink": 0.5294117647058824, "brown": 0.17647058823529413}, "focus": {"false": 0.47058823529411764, "true": 0.5294117647058824}, "group": {"0": 0.29411764705882354, "1": 0.5294117647058824, "2": 0.17647058823529413}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"3/17\"],[\"brown\",\"3/17\"],[\"pink\",\"9/17\"],[\"purple\",\"2/17\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.96, "black": 0.01, "purple": 0.01, "brown": 0.02}, "focus": {"false": 0.52, "true": 0.48}, "group": {"0": 0.01, "1": 0.98, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:52", "state_id": "known_chance_v2:test:52", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: red=4, black=6, blue=7, white=2, yellow=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "red": "The sampled ball is red", "white": "The sampled ball is white", "black": "The sampled ball is black", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: red, white", "The drawn color is one of: black, yellow", "The drawn color is one of: blue"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.35, "red": 0.2, "white": 0.1, "black": 0.3, "yellow": 0.05}, "focus": {"false": 0.95, "true": 0.05}, "group": {"0": 0.30000000000000004, "1": 0.35, "2": 0.35}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"3/10\"],[\"blue\",\"7/20\"],[\"red\",\"1/5\"],[\"white\",\"1/10\"],[\"yellow\",\"1/20\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"black": 0.14, "white": 0, "blue": 0.79, "yellow": 0.01, "red": 0.06}, "focus": {"false": 0.92, "true": 0.08}, "group": {"0": 0.03, "1": 0.37, "2": 0.6}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:48", "state_id": "known_chance_v2:test:48", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: brown=4, purple=5, yellow=8, orange=8, cyan=8, blue=4, red=1, white=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"white": "The sampled ball is white", "blue": "The sampled ball is blue", "cyan": "The sampled ball is cyan", "yellow": "The sampled ball is yellow", "brown": "The sampled ball is brown", "purple": "The sampled ball is purple", "red": "The sampled ball is red", "orange": "The sampled ball is orange"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a white ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: brown, orange, red", "The drawn color is one of: purple, cyan, white", "The drawn color is one of: yellow, blue"]}}, "gold": {}, "gold_probs": {"outcome": {"white": 0.02564102564102564, "blue": 0.10256410256410256, "cyan": 0.20512820512820512, "yellow": 0.20512820512820512, "brown": 0.10256410256410256, "purple": 0.1282051282051282, "red": 0.02564102564102564, "orange": 0.20512820512820512}, "focus": {"false": 0.9743589743589743, "true": 0.02564102564102564}, "group": {"0": 0.3333333333333333, "1": 0.358974358974359, "2": 0.3076923076923077}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"4/39\"],[\"brown\",\"4/39\"],[\"cyan\",\"8/39\"],[\"orange\",\"8/39\"],[\"purple\",\"5/39\"],[\"red\",\"1/39\"],[\"white\",\"1/39\"],[\"yellow\",\"8/39\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"brown": 0.05, "blue": 0.01, "red": 0.02, "orange": 0.16, "yellow": 0.64, "white": 0.01, "cyan": 0.08, "purple": 0.03}, "focus": {"false": 0.94, "true": 0.06}, "group": {"0": 0.39, "1": 0.3, "2": 0.31}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:55", "state_id": "known_chance_v2:test:55", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: yellow=4, white=5, cyan=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"white": "The sampled ball is white", "yellow": "The sampled ball is yellow", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: yellow", "The drawn color is one of: white", "The drawn color is one of: cyan"]}}, "gold": {}, "gold_probs": {"outcome": {"white": 0.29411764705882354, "yellow": 0.23529411764705882, "cyan": 0.47058823529411764}, "focus": {"false": 0.5294117647058824, "true": 0.47058823529411764}, "group": {"0": 0.23529411764705882, "1": 0.29411764705882354, "2": 0.47058823529411764}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"8/17\"],[\"white\",\"5/17\"],[\"yellow\",\"4/17\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"yellow": 0.16, "cyan": 0.8, "white": 0.04}, "focus": {"false": 0.65, "true": 0.35}, "group": {"0": 0.02, "1": 0.02, "2": 0.96}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:53", "state_id": "known_chance_v2:test:53", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: black=9, gray=3, pink=9, cyan=3, orange=8, green=4, white=7, yellow=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"gray": "The sampled ball is gray", "cyan": "The sampled ball is cyan", "pink": "The sampled ball is pink", "yellow": "The sampled ball is yellow", "green": "The sampled ball is green", "white": "The sampled ball is white", "black": "The sampled ball is black", "orange": "The sampled ball is orange"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a green ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: black, cyan, white", "The drawn color is one of: gray, orange, yellow", "The drawn color is one of: pink, green"]}}, "gold": {}, "gold_probs": {"outcome": {"gray": 0.06, "cyan": 0.06, "pink": 0.18, "yellow": 0.14, "green": 0.08, "white": 0.14, "black": 0.18, "orange": 0.16}, "focus": {"false": 0.92, "true": 0.08}, "group": {"0": 0.38, "1": 0.36, "2": 0.26}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"9/50\"],[\"cyan\",\"3/50\"],[\"gray\",\"3/50\"],[\"green\",\"2/25\"],[\"orange\",\"4/25\"],[\"pink\",\"9/50\"],[\"white\",\"7/50\"],[\"yellow\",\"7/50\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.08, "black": 0.81, "cyan": 0, "green": 0, "gray": 0.01, "yellow": 0.01, "white": 0.01, "orange": 0.07}, "focus": {"false": 0.89, "true": 0.11}, "group": {"0": 0.73, "1": 0.22, "2": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:57", "state_id": "known_chance_v2:test:57", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: purple=5, yellow=7, orange=4, green=6, cyan=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "purple": "The sampled ball is purple", "orange": "The sampled ball is orange", "yellow": "The sampled ball is yellow", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a purple ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: purple, green", "The drawn color is one of: yellow, cyan", "The drawn color is one of: orange"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.2222222222222222, "purple": 0.18518518518518517, "orange": 0.14814814814814814, "yellow": 0.25925925925925924, "cyan": 0.18518518518518517}, "focus": {"false": 0.8148148148148149, "true": 0.18518518518518517}, "group": {"0": 0.4074074074074074, "1": 0.4444444444444444, "2": 0.14814814814814814}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"5/27\"],[\"green\",\"2/9\"],[\"orange\",\"4/27\"],[\"purple\",\"5/27\"],[\"yellow\",\"7/27\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0.03, "orange": 0.01, "cyan": 0, "purple": 0.05, "yellow": 0.91}, "focus": {"false": 0.8200000000000001, "true": 0.18}, "group": {"0": 0.12, "1": 0.85, "2": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:56", "state_id": "known_chance_v2:test:56", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: brown=4, green=7, cyan=9, gray=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "cyan": "The sampled ball is cyan", "gray": "The sampled ball is gray", "brown": "The sampled ball is brown"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: brown, gray", "The drawn color is one of: green", "The drawn color is one of: cyan"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.2916666666666667, "cyan": 0.375, "gray": 0.16666666666666666, "brown": 0.16666666666666666}, "focus": {"false": 0.625, "true": 0.375}, "group": {"0": 0.3333333333333333, "1": 0.2916666666666667, "2": 0.375}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"1/6\"],[\"cyan\",\"3/8\"],[\"gray\",\"1/6\"],[\"green\",\"7/24\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0.93, "green": 0.02, "brown": 0.04, "gray": 0}, "focus": {"false": 0.6699999999999999, "true": 0.33}, "group": {"0": 0.03, "1": 0.04, "2": 0.93}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:58", "state_id": "known_chance_v2:test:58", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: pink=4, brown=4, black=3, green=9, orange=3, purple=1, white=3, red=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"red": "The sampled ball is red", "black": "The sampled ball is black", "green": "The sampled ball is green", "pink": "The sampled ball is pink", "purple": "The sampled ball is purple", "orange": "The sampled ball is orange", "white": "The sampled ball is white", "brown": "The sampled ball is brown"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a red ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: pink, green, white", "The drawn color is one of: brown, orange, red", "The drawn color is one of: black, purple"]}}, "gold": {}, "gold_probs": {"outcome": {"red": 0.18181818181818182, "black": 0.09090909090909091, "green": 0.2727272727272727, "pink": 0.12121212121212122, "purple": 0.030303030303030304, "orange": 0.09090909090909091, "white": 0.09090909090909091, "brown": 0.12121212121212122}, "focus": {"false": 0.8181818181818181, "true": 0.18181818181818182}, "group": {"0": 0.48484848484848486, "1": 0.3939393939393939, "2": 0.12121212121212122}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/11\"],[\"brown\",\"4/33\"],[\"green\",\"3/11\"],[\"orange\",\"1/11\"],[\"pink\",\"4/33\"],[\"purple\",\"1/33\"],[\"red\",\"2/11\"],[\"white\",\"1/11\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"black": 0, "red": 0.01, "white": 0, "orange": 0, "green": 0.94, "pink": 0.03, "purple": 0.02, "brown": 0}, "focus": {"false": 0.83, "true": 0.17}, "group": {"0": 0.88, "1": 0.1, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:54", "state_id": "known_chance_v2:test:54", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: pink=3, white=1, purple=8, green=9, cyan=8, blue=6, yellow=8, brown=3, gray=5, orange=1, red=9, black=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"red": "The sampled ball is red", "green": "The sampled ball is green", "brown": "The sampled ball is brown", "blue": "The sampled ball is blue", "black": "The sampled ball is black", "orange": "The sampled ball is orange", "purple": "The sampled ball is purple", "white": "The sampled ball is white", "pink": "The sampled ball is pink", "gray": "The sampled ball is gray", "yellow": "The sampled ball is yellow", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: pink, green, yellow, orange", "The drawn color is one of: white, cyan, brown, red", "The drawn color is one of: purple, blue, gray, black"]}}, "gold": {}, "gold_probs": {"outcome": {"red": 0.13636363636363635, "green": 0.13636363636363635, "brown": 0.045454545454545456, "blue": 0.09090909090909091, "black": 0.07575757575757576, "orange": 0.015151515151515152, "purple": 0.12121212121212122, "white": 0.015151515151515152, "pink": 0.045454545454545456, "gray": 0.07575757575757576, "yellow": 0.12121212121212122, "cyan": 0.12121212121212122}, "focus": {"false": 0.8787878787878788, "true": 0.12121212121212122}, "group": {"0": 0.3181818181818182, "1": 0.3181818181818182, "2": 0.36363636363636365}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"5/66\"],[\"blue\",\"1/11\"],[\"brown\",\"1/22\"],[\"cyan\",\"4/33\"],[\"gray\",\"5/66\"],[\"green\",\"3/22\"],[\"orange\",\"1/66\"],[\"pink\",\"1/22\"],[\"purple\",\"4/33\"],[\"red\",\"3/22\"],[\"white\",\"1/66\"],[\"yellow\",\"4/33\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0.02, "brown": 0.01, "purple": 0.06, "orange": 0.03, "black": 0.01, "pink": 0.09, "yellow": 0.02, "gray": 0.01, "cyan": 0.01, "blue": 0.02, "green": 0.5, "red": 0.22}, "focus": {"false": 0.85, "true": 0.15}, "group": {"0": 0.32, "1": 0.24, "2": 0.44}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:61", "state_id": "known_chance_v2:test:61", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: purple=1, black=5, gray=6, orange=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"gray": "The sampled ball is gray", "black": "The sampled ball is black", "orange": "The sampled ball is orange", "purple": "The sampled ball is purple"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: purple, orange", "The drawn color is one of: black", "The drawn color is one of: gray"]}}, "gold": {}, "gold_probs": {"outcome": {"gray": 0.375, "black": 0.3125, "orange": 0.25, "purple": 0.0625}, "focus": {"false": 0.6875, "true": 0.3125}, "group": {"0": 0.3125, "1": 0.3125, "2": 0.375}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"5/16\"],[\"gray\",\"3/8\"],[\"orange\",\"1/4\"],[\"purple\",\"1/16\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"gray": 0.88, "purple": 0.06, "orange": 0.01, "black": 0.05}, "focus": {"false": 0.73, "true": 0.27}, "group": {"0": 0.02, "1": 0.04, "2": 0.94}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:62", "state_id": "known_chance_v2:test:62", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: green=6, cyan=6, orange=6, black=4, purple=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "black": "The sampled ball is black", "green": "The sampled ball is green", "purple": "The sampled ball is purple", "orange": "The sampled ball is orange"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a green ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: green, black", "The drawn color is one of: cyan, purple", "The drawn color is one of: orange"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.21428571428571427, "black": 0.14285714285714285, "green": 0.21428571428571427, "purple": 0.21428571428571427, "orange": 0.21428571428571427}, "focus": {"false": 0.7857142857142857, "true": 0.21428571428571427}, "group": {"0": 0.3571428571428571, "1": 0.42857142857142855, "2": 0.21428571428571427}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/7\"],[\"cyan\",\"3/14\"],[\"green\",\"3/14\"],[\"orange\",\"3/14\"],[\"purple\",\"3/14\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"orange": 0.02, "purple": 0.01, "cyan": 0.02, "green": 0.68, "black": 0.27}, "focus": {"false": 0.77, "true": 0.23}, "group": {"0": 0.61, "1": 0.31, "2": 0.08}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:59", "state_id": "known_chance_v2:test:59", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: black=1, white=8, cyan=6, purple=4, yellow=7, green=4, brown=9, gray=7, pink=7, orange=6, red=1, blue=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"purple": "The sampled ball is purple", "orange": "The sampled ball is orange", "blue": "The sampled ball is blue", "red": "The sampled ball is red", "yellow": "The sampled ball is yellow", "green": "The sampled ball is green", "black": "The sampled ball is black", "cyan": "The sampled ball is cyan", "white": "The sampled ball is white", "gray": "The sampled ball is gray", "pink": "The sampled ball is pink", "brown": "The sampled ball is brown"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: black, purple, brown, orange", "The drawn color is one of: white, yellow, gray, red", "The drawn color is one of: cyan, green, pink, blue"]}}, "gold": {}, "gold_probs": {"outcome": {"purple": 0.05970149253731343, "orange": 0.08955223880597014, "blue": 0.1044776119402985, "red": 0.014925373134328358, "yellow": 0.1044776119402985, "green": 0.05970149253731343, "black": 0.014925373134328358, "cyan": 0.08955223880597014, "white": 0.11940298507462686, "gray": 0.1044776119402985, "pink": 0.1044776119402985, "brown": 0.13432835820895522}, "focus": {"false": 0.8955223880597015, "true": 0.1044776119402985}, "group": {"0": 0.29850746268656714, "1": 0.34328358208955223, "2": 0.3582089552238806}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/67\"],[\"blue\",\"7/67\"],[\"brown\",\"9/67\"],[\"cyan\",\"6/67\"],[\"gray\",\"7/67\"],[\"green\",\"4/67\"],[\"orange\",\"6/67\"],[\"pink\",\"7/67\"],[\"purple\",\"4/67\"],[\"red\",\"1/67\"],[\"white\",\"8/67\"],[\"yellow\",\"7/67\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"brown": 0.79, "yellow": 0.01, "cyan": 0, "white": 0.13, "purple": 0, "gray": 0, "black": 0.05, "blue": 0.01, "green": 0, "red": 0.01, "pink": 0, "orange": 0}, "focus": {"false": 0.84, "true": 0.16}, "group": {"0": 0.18, "1": 0.45999999999999996, "2": 0.36}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:63", "state_id": "known_chance_v2:test:63", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: pink=7, white=7, blue=1, orange=2, green=6, purple=1, gray=2, black=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"white": "The sampled ball is white", "green": "The sampled ball is green", "purple": "The sampled ball is purple", "blue": "The sampled ball is blue", "black": "The sampled ball is black", "orange": "The sampled ball is orange", "pink": "The sampled ball is pink", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a green ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: pink, orange, gray", "The drawn color is one of: white, green, black", "The drawn color is one of: blue, purple"]}}, "gold": {}, "gold_probs": {"outcome": {"white": 0.20588235294117646, "green": 0.17647058823529413, "purple": 0.029411764705882353, "blue": 0.029411764705882353, "black": 0.23529411764705882, "orange": 0.058823529411764705, "pink": 0.20588235294117646, "gray": 0.058823529411764705}, "focus": {"false": 0.8235294117647058, "true": 0.17647058823529413}, "group": {"0": 0.32352941176470584, "1": 0.6176470588235294, "2": 0.058823529411764705}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"4/17\"],[\"blue\",\"1/34\"],[\"gray\",\"1/17\"],[\"green\",\"3/17\"],[\"orange\",\"1/17\"],[\"pink\",\"7/34\"],[\"purple\",\"1/34\"],[\"white\",\"7/34\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0.01, "orange": 0, "blue": 0, "pink": 0.03, "gray": 0, "white": 0.01, "black": 0.95, "purple": 0}, "focus": {"false": 0.84, "true": 0.16}, "group": {"0": 0.01, "1": 0.98, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:ood:0", "state_id": "smart_home_v2:ood:0", "family_id": "smart_home_v2", "split": "ood", "state": "家庭控制记录：房间=office，设备=fan，动作=off；已授权=yes，人数=0，小时=0。", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "office fan", "unit_1": "garage speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"office\",\"fan\",true,0]", "template_id": "smart_home_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:ood:1", "state_id": "smart_home_v2:ood:1", "family_id": "smart_home_v2", "split": "ood", "state": "家庭控制记录：房间=bathroom，设备=heater，动作=off；已授权=yes，人数=0，小时=10。", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bathroom heater", "unit_1": "bedroom speaker", "unit_2": "attic heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 1}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 1, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bathroom\",\"heater\",true,0]", "template_id": "smart_home_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_2": 0, "unit_1": 0}, "execute": {"false": 0.96, "true": 0.04}, "risk": {"0": 0, "1": 1, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:test:60", "state_id": "known_chance_v2:test:60", "family_id": "known_chance_v2", "split": "test", "state": "Unobserved trial: exactly one ball is drawn uniformly from a bag. Counts: gray=1, yellow=2, brown=8. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"yellow": "The sampled ball is yellow", "brown": "The sampled ball is brown", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a gray ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: gray", "The drawn color is one of: yellow", "The drawn color is one of: brown"]}}, "gold": {}, "gold_probs": {"outcome": {"yellow": 0.18181818181818182, "brown": 0.7272727272727273, "gray": 0.09090909090909091}, "focus": {"false": 0.9090909090909091, "true": 0.09090909090909091}, "group": {"0": 0.09090909090909091, "1": 0.18181818181818182, "2": 0.7272727272727273}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"8/11\"],[\"gray\",\"1/11\"],[\"yellow\",\"2/11\"]]", "template_id": "known_chance_v2:test"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"gray": 0.02, "yellow": 0, "brown": 0.98}, "focus": {"false": 0.9, "true": 0.1}, "group": {"0": 0.01, "1": 0, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:ood:2", "state_id": "smart_home_v2:ood:2", "family_id": "smart_home_v2", "split": "ood", "state": "家庭控制记录：房间=living room，设备=speaker，动作=on；已授权=no，人数=3，小时=11。", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garden lamp", "unit_1": "office lamp", "unit_2": "hall fan", "unit_3": "living room speaker", "unit_4": "bedroom fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_3", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 1, "unit_4": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"living room\",\"speaker\",false,3]", "template_id": "smart_home_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_2": 0, "unit_1": 0, "unit_4": 0, "unit_3": 1}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:ood:4", "state_id": "smart_home_v2:ood:4", "family_id": "smart_home_v2", "split": "ood", "state": "家庭控制记录：房间=library，设备=fan，动作=on；已授权=yes，人数=0，小时=23。", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library heater", "unit_1": "living room lamp", "unit_2": "garden heater", "unit_3": "library speaker", "unit_4": "office lamp", "unit_5": "attic lamp", "unit_6": "bedroom fan", "unit_7": "bedroom heater", "unit_8": "kitchen speaker", "unit_9": "bathroom lamp", "unit_10": "living room speaker", "unit_11": "library fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_11", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"library\",\"fan\",true,0]", "template_id": "smart_home_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_3": 0, "unit_10": 0, "unit_9": 0, "unit_11": 1, "unit_0": 0, "unit_5": 0, "unit_4": 0, "unit_8": 0, "unit_7": 0, "unit_1": 0, "unit_6": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:ood:5", "state_id": "smart_home_v2:ood:5", "family_id": "smart_home_v2", "split": "ood", "state": "家庭控制记录：房间=bedroom，设备=fan，动作=off；已授权=no，人数=1，小时=14。", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "kitchen speaker", "unit_1": "bedroom fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bedroom\",\"fan\",false,1]", "template_id": "smart_home_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 1, "unit_0": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:ood:3", "state_id": "smart_home_v2:ood:3", "family_id": "smart_home_v2", "split": "ood", "state": "家庭控制记录：房间=garden，设备=speaker，动作=off；已授权=yes，人数=2，小时=17。", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "kitchen lamp", "unit_1": "bathroom speaker", "unit_2": "garden speaker", "unit_3": "hall fan", "unit_4": "bedroom lamp", "unit_5": "living room fan", "unit_6": "bathroom lamp", "unit_7": "kitchen speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garden\",\"speaker\",true,2]", "template_id": "smart_home_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_7": 0, "unit_2": 1, "unit_6": 0, "unit_3": 0, "unit_5": 0, "unit_1": 0, "unit_4": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:ood:8", "state_id": "smart_home_v2:ood:8", "family_id": "smart_home_v2", "split": "ood", "state": "家庭控制记录：房间=dining room，设备=lamp，动作=on；已授权=no，人数=1，小时=22。", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "attic fan", "unit_1": "office fan", "unit_2": "garage lamp", "unit_3": "garage heater", "unit_4": "bedroom fan", "unit_5": "bathroom speaker", "unit_6": "dining room lamp", "unit_7": "office lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_6", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 1, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"dining room\",\"lamp\",false,1]", "template_id": "smart_home_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_7": 0, "unit_4": 0, "unit_5": 0, "unit_2": 0, "unit_0": 0, "unit_6": 1, "unit_3": 0}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:ood:6", "state_id": "smart_home_v2:ood:6", "family_id": "smart_home_v2", "split": "ood", "state": "家庭控制记录：房间=hall，设备=heater，动作=off；已授权=yes，人数=1，小时=13。", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "hall heater", "unit_1": "balcony fan", "unit_2": "bedroom lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"hall\",\"heater\",true,1]", "template_id": "smart_home_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_2": 0, "unit_1": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:ood:7", "state_id": "smart_home_v2:ood:7", "family_id": "smart_home_v2", "split": "ood", "state": "家庭控制记录：房间=kitchen，设备=heater，动作=on；已授权=yes，人数=2，小时=18。", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "kitchen speaker", "unit_1": "hall lamp", "unit_2": "dining room fan", "unit_3": "garden fan", "unit_4": "kitchen heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_4", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"kitchen\",\"heater\",true,2]", "template_id": "smart_home_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_4": 1, "unit_1": 0, "unit_3": 0, "unit_0": 0, "unit_2": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:ood:9", "state_id": "smart_home_v2:ood:9", "family_id": "smart_home_v2", "split": "ood", "state": "家庭控制记录：房间=garage，设备=fan，动作=on；已授权=yes，人数=2，小时=13。", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garden heater", "unit_1": "kitchen fan", "unit_2": "attic speaker", "unit_3": "living room speaker", "unit_4": "garage fan", "unit_5": "garage heater", "unit_6": "attic lamp", "unit_7": "balcony heater", "unit_8": "balcony lamp", "unit_9": "hall lamp", "unit_10": "office fan", "unit_11": "library heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_4", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 1, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garage\",\"fan\",true,2]", "template_id": "smart_home_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_5": 0, "unit_10": 0, "unit_9": 0, "unit_7": 0, "unit_4": 1, "unit_8": 0, "unit_6": 0, "unit_3": 0, "unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_11": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:ood:10", "state_id": "smart_home_v2:ood:10", "family_id": "smart_home_v2", "split": "ood", "state": "家庭控制记录：房间=bedroom，设备=speaker，动作=off；已授权=yes，人数=2，小时=22。", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom heater", "unit_1": "bedroom speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bedroom\",\"speaker\",true,2]", "template_id": "smart_home_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 1, "unit_0": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:ood:12", "state_id": "smart_home_v2:ood:12", "family_id": "smart_home_v2", "split": "ood", "state": "家庭控制记录：房间=bathroom，设备=lamp，动作=on；已授权=yes，人数=0，小时=17。", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "hall fan", "unit_1": "bathroom lamp", "unit_2": "hall speaker", "unit_3": "living room fan", "unit_4": "library heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0, "unit_3": 0, "unit_4": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bathroom\",\"lamp\",true,0]", "template_id": "smart_home_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_3": 0, "unit_0": 0, "unit_1": 1, "unit_2": 0, "unit_4": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:ood:13", "state_id": "smart_home_v2:ood:13", "family_id": "smart_home_v2", "split": "ood", "state": "家庭控制记录：房间=dining room，设备=heater，动作=on；已授权=no，人数=3，小时=23。", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library speaker", "unit_1": "attic fan", "unit_2": "library heater", "unit_3": "garden speaker", "unit_4": "garden heater", "unit_5": "hall lamp", "unit_6": "garden lamp", "unit_7": "dining room heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_7", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"dining room\",\"heater\",false,3]", "template_id": "smart_home_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_6": 0, "unit_4": 0, "unit_7": 1, "unit_2": 0, "unit_0": 0, "unit_5": 0, "unit_3": 0, "unit_1": 0}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:ood:11", "state_id": "smart_home_v2:ood:11", "family_id": "smart_home_v2", "split": "ood", "state": "家庭控制记录：房间=dining room，设备=fan，动作=on；已授权=no，人数=1，小时=8。", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "hall fan", "unit_1": "dining room fan", "unit_2": "living room fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"dining room\",\"fan\",false,1]", "template_id": "smart_home_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_0": 0, "unit_1": 1}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:ood:17", "state_id": "smart_home_v2:ood:17", "family_id": "smart_home_v2", "split": "ood", "state": "家庭控制记录：房间=bathroom，设备=speaker，动作=on；已授权=no，人数=3，小时=2。", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "attic heater", "unit_1": "balcony heater", "unit_2": "bathroom lamp", "unit_3": "office heater", "unit_4": "bathroom speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_4", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bathroom\",\"speaker\",false,3]", "template_id": "smart_home_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 0, "unit_2": 0, "unit_4": 1, "unit_1": 0, "unit_3": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:ood:14", "state_id": "smart_home_v2:ood:14", "family_id": "smart_home_v2", "split": "ood", "state": "家庭控制记录：房间=hall，设备=lamp，动作=on；已授权=no，人数=1，小时=16。", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "hall heater", "unit_1": "library heater", "unit_2": "living room heater", "unit_3": "office lamp", "unit_4": "bedroom fan", "unit_5": "balcony lamp", "unit_6": "kitchen heater", "unit_7": "bedroom speaker", "unit_8": "garden speaker", "unit_9": "hall lamp", "unit_10": "attic fan", "unit_11": "attic speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_9", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 1, "unit_10": 0, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"hall\",\"lamp\",false,1]", "template_id": "smart_home_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_8": 0, "unit_11": 0, "unit_4": 0, "unit_3": 0, "unit_9": 1, "unit_7": 0, "unit_1": 0, "unit_6": 0, "unit_10": 0, "unit_0": 0, "unit_5": 0}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:ood:18", "state_id": "smart_home_v2:ood:18", "family_id": "smart_home_v2", "split": "ood", "state": "家庭控制记录：房间=balcony，设备=lamp，动作=off；已授权=yes，人数=0，小时=9。", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "dining room fan", "unit_1": "balcony lamp", "unit_2": "attic lamp", "unit_3": "balcony fan", "unit_4": "office fan", "unit_5": "garden lamp", "unit_6": "hall fan", "unit_7": "bathroom lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"balcony\",\"lamp\",true,0]", "template_id": "smart_home_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_3": 0, "unit_6": 0, "unit_0": 0, "unit_5": 0, "unit_4": 0, "unit_1": 1, "unit_7": 0}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:ood:19", "state_id": "smart_home_v2:ood:19", "family_id": "smart_home_v2", "split": "ood", "state": "家庭控制记录：房间=living room，设备=fan，动作=off；已授权=yes，人数=1，小时=18。", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "hall speaker", "unit_1": "balcony heater", "unit_2": "kitchen heater", "unit_3": "hall heater", "unit_4": "garage fan", "unit_5": "attic fan", "unit_6": "garage speaker", "unit_7": "balcony speaker", "unit_8": "dining room fan", "unit_9": "kitchen lamp", "unit_10": "garage lamp", "unit_11": "living room fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_11", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 1}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"living room\",\"fan\",true,1]", "template_id": "smart_home_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_10": 0, "unit_4": 0, "unit_6": 0, "unit_8": 0, "unit_1": 0, "unit_3": 0, "unit_5": 0, "unit_7": 0, "unit_9": 0, "unit_0": 0, "unit_11": 1}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:ood:16", "state_id": "smart_home_v2:ood:16", "family_id": "smart_home_v2", "split": "ood", "state": "家庭控制记录：房间=living room，设备=fan，动作=off；已授权=yes，人数=2，小时=9。", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room heater", "unit_1": "living room fan", "unit_2": "attic heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_1", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 1, "unit_2": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"living room\",\"fan\",true,2]", "template_id": "smart_home_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 1, "unit_2": 0, "unit_0": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:ood:15", "state_id": "smart_home_v2:ood:15", "family_id": "smart_home_v2", "split": "ood", "state": "家庭控制记录：房间=bedroom，设备=heater，动作=off；已授权=yes，人数=1，小时=2。", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bedroom heater", "unit_1": "library fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bedroom\",\"heater\",true,1]", "template_id": "smart_home_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_0": 1}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:ood:21", "state_id": "smart_home_v2:ood:21", "family_id": "smart_home_v2", "split": "ood", "state": "家庭控制记录：房间=library，设备=lamp，动作=off；已授权=no，人数=3，小时=13。", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library lamp", "unit_1": "library heater", "unit_2": "bathroom lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"library\",\"lamp\",false,3]", "template_id": "smart_home_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_1": 0, "unit_0": 1, "unit_2": 0}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:ood:20", "state_id": "smart_home_v2:ood:20", "family_id": "smart_home_v2", "split": "ood", "state": "家庭控制记录：房间=bathroom，设备=fan，动作=on；已授权=yes，人数=3，小时=15。", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "bathroom fan", "unit_1": "balcony heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bathroom\",\"fan\",true,3]", "template_id": "smart_home_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:ood:22", "state_id": "smart_home_v2:ood:22", "family_id": "smart_home_v2", "split": "ood", "state": "家庭控制记录：房间=garden，设备=lamp，动作=off；已授权=no，人数=1，小时=10。", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "garage heater", "unit_1": "library heater", "unit_2": "garden lamp", "unit_3": "kitchen speaker", "unit_4": "balcony lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1, "unit_3": 0, "unit_4": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"garden\",\"lamp\",false,1]", "template_id": "smart_home_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 1, "unit_3": 0, "unit_1": 0, "unit_0": 0, "unit_4": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:ood:24", "state_id": "smart_home_v2:ood:24", "family_id": "smart_home_v2", "split": "ood", "state": "家庭控制记录：房间=dining room，设备=heater，动作=off；已授权=yes，人数=0，小时=1。", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "living room fan", "unit_1": "balcony heater", "unit_2": "garage heater", "unit_3": "library fan", "unit_4": "kitchen fan", "unit_5": "bathroom fan", "unit_6": "library heater", "unit_7": "dining room heater", "unit_8": "hall heater", "unit_9": "bedroom lamp", "unit_10": "garage speaker", "unit_11": "living room speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_7", "execute": false, "risk": 1}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 1, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 1, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"dining room\",\"heater\",true,0]", "template_id": "smart_home_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_7": 1, "unit_9": 0, "unit_10": 0, "unit_2": 0, "unit_0": 0, "unit_4": 0, "unit_1": 0, "unit_5": 0, "unit_8": 0, "unit_6": 0, "unit_3": 0, "unit_11": 0}, "execute": {"false": 0.96, "true": 0.04}, "risk": {"0": 0, "1": 1, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:ood:23", "state_id": "smart_home_v2:ood:23", "family_id": "smart_home_v2", "split": "ood", "state": "家庭控制记录：房间=attic，设备=lamp，动作=on；已授权=yes，人数=2，小时=22。", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "dining room lamp", "unit_1": "balcony speaker", "unit_2": "living room lamp", "unit_3": "office fan", "unit_4": "balcony lamp", "unit_5": "attic lamp", "unit_6": "hall lamp", "unit_7": "library fan"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_5", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 1, "unit_6": 0, "unit_7": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"attic\",\"lamp\",true,2]", "template_id": "smart_home_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_3": 0, "unit_4": 0, "unit_2": 0, "unit_0": 0, "unit_6": 0, "unit_7": 0, "unit_5": 1, "unit_1": 0}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:ood:25", "state_id": "smart_home_v2:ood:25", "family_id": "smart_home_v2", "split": "ood", "state": "家庭控制记录：房间=library，设备=fan，动作=off；已授权=no，人数=3，小时=0。", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "library fan", "unit_1": "balcony lamp"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"library\",\"fan\",false,3]", "template_id": "smart_home_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:ood:28", "state_id": "smart_home_v2:ood:28", "family_id": "smart_home_v2", "split": "ood", "state": "家庭控制记录：房间=balcony，设备=heater，动作=off；已授权=no，人数=0，小时=12。", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "dining room speaker", "unit_1": "attic heater", "unit_2": "bedroom fan", "unit_3": "balcony heater", "unit_4": "garage fan", "unit_5": "garden lamp", "unit_6": "bathroom heater", "unit_7": "garden speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_3", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 1, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"balcony\",\"heater\",false,0]", "template_id": "smart_home_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_3": 1, "unit_6": 0, "unit_2": 0, "unit_0": 0, "unit_4": 0, "unit_5": 0, "unit_1": 0, "unit_7": 0}, "execute": {"false": 0.99, "true": 0.01}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:ood:26", "state_id": "smart_home_v2:ood:26", "family_id": "smart_home_v2", "split": "ood", "state": "家庭控制记录：房间=attic，设备=lamp，动作=off；已授权=yes，人数=1，小时=23。", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "attic lamp", "unit_1": "garden speaker", "unit_2": "kitchen speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"attic\",\"lamp\",true,1]", "template_id": "smart_home_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_1": 0, "unit_0": 1}, "execute": {"false": 0.020000000000000018, "true": 0.98}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:ood:27", "state_id": "smart_home_v2:ood:27", "family_id": "smart_home_v2", "split": "ood", "state": "家庭控制记录：房间=library，设备=heater，动作=off；已授权=no，人数=1，小时=14。", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "kitchen lamp", "unit_1": "dining room fan", "unit_2": "library heater", "unit_3": "library fan", "unit_4": "attic heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_2", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 1, "unit_3": 0, "unit_4": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"library\",\"heater\",false,1]", "template_id": "smart_home_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_3": 0, "unit_2": 1, "unit_4": 0, "unit_0": 0, "unit_1": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:ood:29", "state_id": "smart_home_v2:ood:29", "family_id": "smart_home_v2", "split": "ood", "state": "家庭控制记录：房间=bathroom，设备=heater，动作=on；已授权=no，人数=2，小时=14。", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "office fan", "unit_1": "bedroom fan", "unit_2": "garden heater", "unit_3": "office lamp", "unit_4": "bathroom lamp", "unit_5": "attic fan", "unit_6": "garden fan", "unit_7": "garage lamp", "unit_8": "bathroom speaker", "unit_9": "balcony heater", "unit_10": "office speaker", "unit_11": "bathroom heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_11", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 0, "unit_1": 0, "unit_2": 0, "unit_3": 0, "unit_4": 0, "unit_5": 0, "unit_6": 0, "unit_7": 0, "unit_8": 0, "unit_9": 0, "unit_10": 0, "unit_11": 1}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"bathroom\",\"heater\",false,2]", "template_id": "smart_home_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_11": 1, "unit_0": 0, "unit_1": 0, "unit_4": 0, "unit_8": 0, "unit_10": 0, "unit_2": 0, "unit_5": 0, "unit_9": 0, "unit_6": 0, "unit_7": 0, "unit_3": 0}, "execute": {"false": 0.98, "true": 0.02}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:ood:31", "state_id": "smart_home_v2:ood:31", "family_id": "smart_home_v2", "split": "ood", "state": "家庭控制记录：房间=balcony，设备=fan，动作=on；已授权=yes，人数=0，小时=20。", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "balcony fan", "unit_1": "bedroom speaker", "unit_2": "balcony speaker"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": true, "risk": 0}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0, "unit_2": 0}, "execute": {"false": 0, "true": 1}, "risk": {"0": 1, "1": 0, "2": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"balcony\",\"fan\",true,0]", "template_id": "smart_home_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_2": 0, "unit_1": 0, "unit_0": 1}, "execute": {"false": 0.030000000000000027, "true": 0.97}, "risk": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:ood:0", "state_id": "catalog_lookup_v2:ood:0", "family_id": "catalog_lookup_v2", "split": "ood", "state": "库存记录: item=SKU-00229; color=brown; size=medium; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "green, large", "candidate_1": "brown, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00229\",\"brown\",\"medium\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 0.10999999999999999, "true": 0.89}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "smart_home_v2:ood:30", "state_id": "smart_home_v2:ood:30", "family_id": "smart_home_v2", "split": "ood", "state": "家庭控制记录：房间=office，设备=speaker，动作=off；已授权=no，人数=2，小时=21。", "questions": {"device": {"type": "choice", "instructions": "Select exactly the room and device named in the request. Ignore authorization and occupancy for this question.", "criteria": {"unit_0": "office speaker", "unit_1": "hall heater"}}, "execute": {"type": "boolean", "instructions": "Is execution allowed? Rule: authorization is required for every command; additionally any heater command requires at least one person present. Other devices do not require occupancy."}, "risk": {"type": "score", "instructions": "Classify the control request using these mutually exclusive rules.", "criteria": ["Authorized, and either not a heater or at least one person present", "Authorized heater request with zero people present", "Not authorized, regardless of device or occupancy"]}}, "gold": {"device": "unit_0", "execute": false, "risk": 2}, "gold_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 1, "true": 0}, "risk": {"0": 0, "1": 0, "2": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "home:[\"office\",\"speaker\",false,2]", "template_id": "smart_home_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"device": {"unit_0": 1, "unit_1": 0}, "execute": {"false": 0.97, "true": 0.03}, "risk": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:ood:1", "state_id": "catalog_lookup_v2:ood:1", "family_id": "catalog_lookup_v2", "split": "ood", "state": "库存记录: item=SKU-00144; color=black; size=small; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "black, small", "candidate_1": "pink, small", "candidate_2": "white, large", "candidate_3": "brown, large", "candidate_4": "gray, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00144\",\"black\",\"small\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 1, "candidate_3": 0, "candidate_4": 0, "candidate_2": 0, "candidate_1": 0}, "ready": {"false": 0.77, "true": 0.23}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:ood:2", "state_id": "catalog_lookup_v2:ood:2", "family_id": "catalog_lookup_v2", "split": "ood", "state": "库存记录: item=SKU-00126; color=white; size=small; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "brown, large", "candidate_1": "white, medium", "candidate_2": "yellow, small", "candidate_3": "green, large", "candidate_4": "blue, small", "candidate_5": "white, small", "candidate_6": "yellow, large", "candidate_7": "black, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_5", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 1, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00126\",\"white\",\"small\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_7": 0, "candidate_5": 1, "candidate_3": 0, "candidate_4": 0, "candidate_6": 0, "candidate_2": 0, "candidate_0": 0, "candidate_1": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:ood:3", "state_id": "catalog_lookup_v2:ood:3", "family_id": "catalog_lookup_v2", "split": "ood", "state": "库存记录: item=SKU-00129; color=white; size=medium; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "white, small", "candidate_1": "cyan, medium", "candidate_2": "blue, small", "candidate_3": "cyan, large", "candidate_4": "red, large", "candidate_5": "black, large", "candidate_6": "orange, medium", "candidate_7": "red, small", "candidate_8": "pink, small", "candidate_9": "brown, medium", "candidate_10": "gray, large", "candidate_11": "white, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_11", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00129\",\"white\",\"medium\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_10": 0, "candidate_11": 1, "candidate_5": 0, "candidate_6": 0, "candidate_8": 0, "candidate_2": 0, "candidate_4": 0, "candidate_0": 0, "candidate_7": 0, "candidate_3": 0, "candidate_9": 0}, "ready": {"false": 0.8, "true": 0.2}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:ood:4", "state_id": "catalog_lookup_v2:ood:4", "family_id": "catalog_lookup_v2", "split": "ood", "state": "库存记录: item=SKU-00111; color=purple; size=medium; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "blue, large", "candidate_1": "white, medium", "candidate_2": "gray, small", "candidate_3": "green, medium", "candidate_4": "black, large", "candidate_5": "green, large", "candidate_6": "purple, medium", "candidate_7": "brown, large", "candidate_8": "white, small", "candidate_9": "gray, medium", "candidate_10": "black, medium", "candidate_11": "brown, small", "candidate_12": "yellow, small", "candidate_13": "pink, medium", "candidate_14": "pink, small", "candidate_15": "red, small", "candidate_16": "yellow, large", "candidate_17": "blue, small", "candidate_18": "gray, large", "candidate_19": "orange, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_6", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 1, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00111\",\"purple\",\"medium\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_13": 0, "candidate_9": 0, "candidate_7": 0, "candidate_18": 0, "candidate_12": 0, "candidate_11": 0, "candidate_10": 0, "candidate_2": 0, "candidate_5": 0, "candidate_0": 0, "candidate_19": 0, "candidate_1": 0, "candidate_4": 0, "candidate_6": 1, "candidate_17": 0, "candidate_8": 0, "candidate_3": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:ood:5", "state_id": "catalog_lookup_v2:ood:5", "family_id": "catalog_lookup_v2", "split": "ood", "state": "库存记录: item=SKU-00044; color=blue; size=large; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "blue, large", "candidate_1": "red, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00044\",\"blue\",\"large\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_0": 1}, "ready": {"false": 0.16000000000000003, "true": 0.84}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:ood:6", "state_id": "catalog_lookup_v2:ood:6", "family_id": "catalog_lookup_v2", "split": "ood", "state": "库存记录: item=SKU-00231; color=brown; size=medium; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "white, large", "candidate_1": "brown, medium", "candidate_2": "gray, large", "candidate_3": "green, medium", "candidate_4": "blue, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00231\",\"brown\",\"medium\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_4": 0, "candidate_2": 0, "candidate_0": 0, "candidate_3": 0, "candidate_1": 1}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:ood:7", "state_id": "catalog_lookup_v2:ood:7", "family_id": "catalog_lookup_v2", "split": "ood", "state": "库存记录: item=SKU-00163; color=black; size=large; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "yellow, medium", "candidate_1": "red, large", "candidate_2": "cyan, medium", "candidate_3": "white, medium", "candidate_4": "yellow, small", "candidate_5": "gray, medium", "candidate_6": "green, large", "candidate_7": "black, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_7", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00163\",\"black\",\"large\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_5": 0, "candidate_1": 0, "candidate_2": 0, "candidate_6": 0, "candidate_0": 0, "candidate_7": 1, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 0.9299999999999999, "true": 0.07}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:ood:9", "state_id": "catalog_lookup_v2:ood:9", "family_id": "catalog_lookup_v2", "split": "ood", "state": "库存记录: item=SKU-00051; color=green; size=small; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "brown, large", "candidate_1": "black, large", "candidate_2": "pink, small", "candidate_3": "blue, large", "candidate_4": "yellow, small", "candidate_5": "black, small", "candidate_6": "yellow, medium", "candidate_7": "purple, small", "candidate_8": "blue, medium", "candidate_9": "black, medium", "candidate_10": "pink, medium", "candidate_11": "purple, large", "candidate_12": "green, medium", "candidate_13": "brown, medium", "candidate_14": "orange, medium", "candidate_15": "white, small", "candidate_16": "green, large", "candidate_17": "gray, large", "candidate_18": "red, medium", "candidate_19": "green, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_19", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00051\",\"green\",\"small\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_8": 0, "candidate_18": 0, "candidate_14": 0, "candidate_13": 0, "candidate_7": 0, "candidate_1": 0, "candidate_19": 1, "candidate_6": 0, "candidate_0": 0, "candidate_12": 0, "candidate_2": 0, "candidate_11": 0, "candidate_16": 0, "candidate_4": 0, "candidate_3": 0, "candidate_9": 0, "candidate_10": 0, "candidate_5": 0, "candidate_15": 0, "candidate_17": 0}, "ready": {"false": 0.92, "true": 0.08}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:ood:10", "state_id": "catalog_lookup_v2:ood:10", "family_id": "catalog_lookup_v2", "split": "ood", "state": "库存记录: item=SKU-00175; color=orange; size=small; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "orange, small", "candidate_1": "gray, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00175\",\"orange\",\"small\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:ood:11", "state_id": "catalog_lookup_v2:ood:11", "family_id": "catalog_lookup_v2", "split": "ood", "state": "库存记录: item=SKU-00122; color=white; size=small; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "white, small", "candidate_1": "black, small", "candidate_2": "brown, small", "candidate_3": "yellow, medium", "candidate_4": "blue, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00122\",\"white\",\"small\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_4": 0, "candidate_0": 1, "candidate_3": 0, "candidate_2": 0}, "ready": {"false": 0.91, "true": 0.09}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:ood:12", "state_id": "catalog_lookup_v2:ood:12", "family_id": "catalog_lookup_v2", "split": "ood", "state": "库存记录: item=SKU-00007; color=red; size=small; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "orange, small", "candidate_1": "gray, small", "candidate_2": "red, medium", "candidate_3": "green, large", "candidate_4": "cyan, large", "candidate_5": "red, small", "candidate_6": "brown, large", "candidate_7": "purple, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_5", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 1, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00007\",\"red\",\"small\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_4": 0, "candidate_0": 0, "candidate_3": 0, "candidate_2": 0, "candidate_6": 0, "candidate_1": 0, "candidate_7": 0, "candidate_5": 1}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:ood:13", "state_id": "catalog_lookup_v2:ood:13", "family_id": "catalog_lookup_v2", "split": "ood", "state": "库存记录: item=SKU-00052; color=green; size=small; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "black, large", "candidate_1": "green, large", "candidate_2": "cyan, small", "candidate_3": "pink, medium", "candidate_4": "yellow, medium", "candidate_5": "blue, medium", "candidate_6": "blue, large", "candidate_7": "orange, medium", "candidate_8": "brown, large", "candidate_9": "green, small", "candidate_10": "gray, medium", "candidate_11": "orange, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_9", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 1, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00052\",\"green\",\"small\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_9": 1, "candidate_3": 0, "candidate_0": 0, "candidate_6": 0, "candidate_4": 0, "candidate_1": 0, "candidate_5": 0, "candidate_2": 0, "candidate_10": 0, "candidate_11": 0, "candidate_8": 0, "candidate_7": 0}, "ready": {"false": 0.15000000000000002, "true": 0.85}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:ood:8", "state_id": "catalog_lookup_v2:ood:8", "family_id": "catalog_lookup_v2", "split": "ood", "state": "库存记录: item=SKU-00057; color=green; size=medium; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "white, medium", "candidate_1": "orange, medium", "candidate_2": "cyan, medium", "candidate_3": "white, large", "candidate_4": "brown, large", "candidate_5": "green, large", "candidate_6": "gray, large", "candidate_7": "green, medium", "candidate_8": "orange, small", "candidate_9": "red, large", "candidate_10": "blue, large", "candidate_11": "brown, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_7", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 1, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00057\",\"green\",\"medium\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_5": 0, "candidate_0": 0, "candidate_3": 0, "candidate_8": 0, "candidate_7": 1, "candidate_2": 0, "candidate_4": 0, "candidate_10": 0, "candidate_6": 0, "candidate_9": 0, "candidate_11": 0}, "ready": {"false": 0.8, "true": 0.2}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:ood:15", "state_id": "catalog_lookup_v2:ood:15", "family_id": "catalog_lookup_v2", "split": "ood", "state": "库存记录: item=SKU-00135; color=white; size=medium; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "pink, small", "candidate_1": "white, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00135\",\"white\",\"medium\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:ood:14", "state_id": "catalog_lookup_v2:ood:14", "family_id": "catalog_lookup_v2", "split": "ood", "state": "库存记录: item=SKU-00146; color=black; size=small; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "gray, large", "candidate_1": "orange, medium", "candidate_2": "white, medium", "candidate_3": "gray, small", "candidate_4": "white, small", "candidate_5": "yellow, small", "candidate_6": "green, medium", "candidate_7": "cyan, large", "candidate_8": "brown, large", "candidate_9": "brown, medium", "candidate_10": "white, large", "candidate_11": "black, small", "candidate_12": "black, large", "candidate_13": "yellow, large", "candidate_14": "red, small", "candidate_15": "pink, large", "candidate_16": "yellow, medium", "candidate_17": "orange, large", "candidate_18": "green, small", "candidate_19": "green, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_11", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 1, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00146\",\"black\",\"small\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_17": 0, "candidate_14": 0, "candidate_9": 0, "candidate_0": 0, "candidate_10": 0, "candidate_11": 1, "candidate_19": 0, "candidate_3": 0, "candidate_16": 0, "candidate_8": 0, "candidate_12": 0, "candidate_13": 0, "candidate_4": 0, "candidate_18": 0, "candidate_6": 0, "candidate_15": 0, "candidate_2": 0, "candidate_7": 0, "candidate_1": 0, "candidate_5": 0}, "ready": {"false": 0.89, "true": 0.11}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:ood:16", "state_id": "catalog_lookup_v2:ood:16", "family_id": "catalog_lookup_v2", "split": "ood", "state": "库存记录: item=SKU-00170; color=orange; size=small; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "blue, small", "candidate_1": "brown, large", "candidate_2": "brown, small", "candidate_3": "yellow, large", "candidate_4": "orange, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_4", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00170\",\"orange\",\"small\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 0, "candidate_3": 0, "candidate_2": 0, "candidate_4": 1, "candidate_1": 0}, "ready": {"false": 0.92, "true": 0.08}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:ood:19", "state_id": "catalog_lookup_v2:ood:19", "family_id": "catalog_lookup_v2", "split": "ood", "state": "库存记录: item=SKU-00085; color=yellow; size=medium; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "white, large", "candidate_1": "purple, large", "candidate_2": "purple, medium", "candidate_3": "orange, medium", "candidate_4": "black, small", "candidate_5": "brown, large", "candidate_6": "gray, small", "candidate_7": "black, large", "candidate_8": "pink, small", "candidate_9": "yellow, medium", "candidate_10": "brown, medium", "candidate_11": "yellow, large", "candidate_12": "gray, medium", "candidate_13": "purple, small", "candidate_14": "pink, medium", "candidate_15": "blue, medium", "candidate_16": "blue, small", "candidate_17": "pink, large", "candidate_18": "orange, large", "candidate_19": "blue, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_9", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 1, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00085\",\"yellow\",\"medium\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_18": 0, "candidate_5": 0, "candidate_12": 0, "candidate_15": 0, "candidate_7": 0, "candidate_4": 0, "candidate_3": 0, "candidate_6": 0, "candidate_14": 0, "candidate_13": 0, "candidate_2": 0, "candidate_11": 0, "candidate_17": 0, "candidate_1": 0, "candidate_16": 0, "candidate_0": 0, "candidate_10": 0, "candidate_8": 0, "candidate_19": 0, "candidate_9": 1}, "ready": {"false": 0.10999999999999999, "true": 0.89}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:ood:20", "state_id": "catalog_lookup_v2:ood:20", "family_id": "catalog_lookup_v2", "split": "ood", "state": "库存记录: item=SKU-00120; color=white; size=small; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "gray, large", "candidate_1": "white, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_1", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00120\",\"white\",\"small\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 0, "candidate_1": 1}, "ready": {"false": 0.78, "true": 0.22}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:ood:17", "state_id": "catalog_lookup_v2:ood:17", "family_id": "catalog_lookup_v2", "split": "ood", "state": "库存记录: item=SKU-00043; color=blue; size=large; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "gray, medium", "candidate_1": "pink, large", "candidate_2": "blue, large", "candidate_3": "cyan, large", "candidate_4": "purple, small", "candidate_5": "yellow, large", "candidate_6": "yellow, small", "candidate_7": "gray, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_2", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 1, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00043\",\"blue\",\"large\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_5": 0, "candidate_2": 1, "candidate_6": 0, "candidate_4": 0, "candidate_0": 0, "candidate_7": 0, "candidate_3": 0}, "ready": {"false": 0.9299999999999999, "true": 0.07}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:ood:22", "state_id": "catalog_lookup_v2:ood:22", "family_id": "catalog_lookup_v2", "split": "ood", "state": "库存记录: item=SKU-00156; color=black; size=medium; status=packed; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "green, large", "candidate_1": "orange, medium", "candidate_2": "blue, medium", "candidate_3": "blue, large", "candidate_4": "yellow, medium", "candidate_5": "purple, large", "candidate_6": "gray, medium", "candidate_7": "black, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_7", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 1}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00156\",\"black\",\"medium\",\"packed\",\"yes\"]", "template_id": "catalog_lookup_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_5": 0, "candidate_4": 0, "candidate_7": 1, "candidate_1": 0, "candidate_0": 0, "candidate_3": 0, "candidate_6": 0, "candidate_2": 0}, "ready": {"false": 0.15000000000000002, "true": 0.85}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:ood:21", "state_id": "catalog_lookup_v2:ood:21", "family_id": "catalog_lookup_v2", "split": "ood", "state": "库存记录: item=SKU-00242; color=gray; size=small; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "purple, medium", "candidate_1": "brown, medium", "candidate_2": "gray, small", "candidate_3": "orange, medium", "candidate_4": "yellow, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_2", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 1, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00242\",\"gray\",\"small\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_0": 0, "candidate_2": 1, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 0.92, "true": 0.08}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:ood:24", "state_id": "catalog_lookup_v2:ood:24", "family_id": "catalog_lookup_v2", "split": "ood", "state": "库存记录: item=SKU-00074; color=yellow; size=small; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "brown, large", "candidate_1": "orange, small", "candidate_2": "yellow, small", "candidate_3": "black, large", "candidate_4": "gray, small", "candidate_5": "yellow, medium", "candidate_6": "green, small", "candidate_7": "brown, small", "candidate_8": "blue, small", "candidate_9": "purple, large", "candidate_10": "pink, small", "candidate_11": "white, large", "candidate_12": "black, small", "candidate_13": "pink, medium", "candidate_14": "cyan, medium", "candidate_15": "purple, small", "candidate_16": "black, medium", "candidate_17": "blue, large", "candidate_18": "white, medium", "candidate_19": "red, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_2", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 1, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 0, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00074\",\"yellow\",\"small\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_14": 0, "candidate_18": 0, "candidate_8": 0, "candidate_13": 0, "candidate_7": 0, "candidate_1": 0, "candidate_19": 0, "candidate_12": 0, "candidate_0": 0, "candidate_6": 0, "candidate_2": 1, "candidate_11": 0, "candidate_16": 0, "candidate_4": 0, "candidate_3": 0, "candidate_9": 0, "candidate_10": 0, "candidate_5": 0, "candidate_15": 0, "candidate_17": 0}, "ready": {"false": 0.92, "true": 0.08}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:ood:23", "state_id": "catalog_lookup_v2:ood:23", "family_id": "catalog_lookup_v2", "split": "ood", "state": "库存记录: item=SKU-00018; color=red; size=large; status=picked; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "gray, small", "candidate_1": "white, medium", "candidate_2": "cyan, medium", "candidate_3": "blue, small", "candidate_4": "purple, small", "candidate_5": "red, large", "candidate_6": "brown, small", "candidate_7": "purple, large", "candidate_8": "gray, large", "candidate_9": "yellow, small", "candidate_10": "brown, large", "candidate_11": "yellow, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_5", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 1, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00018\",\"red\",\"large\",\"picked\",\"yes\"]", "template_id": "catalog_lookup_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 0, "candidate_9": 0, "candidate_4": 0, "candidate_6": 0, "candidate_3": 0, "candidate_1": 0, "candidate_10": 0, "candidate_8": 0, "candidate_5": 1, "candidate_11": 0, "candidate_2": 0, "candidate_7": 0}, "ready": {"false": 0.91, "true": 0.09}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:ood:18", "state_id": "catalog_lookup_v2:ood:18", "family_id": "catalog_lookup_v2", "split": "ood", "state": "库存记录: item=SKU-00208; color=pink; size=large; status=new; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "pink, large", "candidate_1": "orange, large", "candidate_2": "blue, large", "candidate_3": "red, medium", "candidate_4": "blue, small", "candidate_5": "white, small", "candidate_6": "brown, medium", "candidate_7": "black, medium", "candidate_8": "blue, medium", "candidate_9": "brown, large", "candidate_10": "green, large", "candidate_11": "purple, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00208\",\"pink\",\"large\",\"new\",\"yes\"]", "template_id": "catalog_lookup_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_3": 0, "candidate_9": 0, "candidate_2": 0, "candidate_8": 0, "candidate_6": 0, "candidate_7": 0, "candidate_1": 0, "candidate_0": 1, "candidate_4": 0, "candidate_11": 0, "candidate_10": 0, "candidate_5": 0}, "ready": {"false": 0.77, "true": 0.23}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:ood:27", "state_id": "catalog_lookup_v2:ood:27", "family_id": "catalog_lookup_v2", "split": "ood", "state": "库存记录: item=SKU-00069; color=green; size=large; status=packed; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "orange, large", "candidate_1": "green, small", "candidate_2": "purple, small", "candidate_3": "cyan, medium", "candidate_4": "red, large", "candidate_5": "green, large", "candidate_6": "blue, large", "candidate_7": "purple, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_5", "ready": true, "stage": 2}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 1, "candidate_6": 0, "candidate_7": 0}, "ready": {"false": 0, "true": 1}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00069\",\"green\",\"large\",\"packed\",\"no\"]", "template_id": "catalog_lookup_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_3": 0, "candidate_2": 0, "candidate_4": 0, "candidate_0": 0, "candidate_1": 0, "candidate_7": 0, "candidate_5": 1, "candidate_6": 0}, "ready": {"false": 0.10999999999999999, "true": 0.89}, "stage": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:ood:26", "state_id": "catalog_lookup_v2:ood:26", "family_id": "catalog_lookup_v2", "split": "ood", "state": "库存记录: item=SKU-00191; color=orange; size=large; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "orange, large", "candidate_1": "purple, medium", "candidate_2": "pink, medium", "candidate_3": "yellow, medium", "candidate_4": "purple, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00191\",\"orange\",\"large\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 1, "candidate_3": 0, "candidate_2": 0, "candidate_4": 0, "candidate_1": 0}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:ood:25", "state_id": "catalog_lookup_v2:ood:25", "family_id": "catalog_lookup_v2", "split": "ood", "state": "库存记录: item=SKU-00019; color=red; size=large; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "red, large", "candidate_1": "gray, small"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00019\",\"red\",\"large\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 0.9299999999999999, "true": 0.07}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:ood:28", "state_id": "catalog_lookup_v2:ood:28", "family_id": "catalog_lookup_v2", "split": "ood", "state": "库存记录: item=SKU-00059; color=green; size=medium; status=picked; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "green, medium", "candidate_1": "yellow, large", "candidate_2": "brown, small", "candidate_3": "brown, medium", "candidate_4": "black, medium", "candidate_5": "red, small", "candidate_6": "purple, medium", "candidate_7": "yellow, small", "candidate_8": "orange, small", "candidate_9": "pink, large", "candidate_10": "green, large", "candidate_11": "red, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 1}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00059\",\"green\",\"medium\",\"picked\",\"no\"]", "template_id": "catalog_lookup_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_9": 0, "candidate_5": 0, "candidate_4": 0, "candidate_0": 1, "candidate_8": 0, "candidate_3": 0, "candidate_1": 0, "candidate_7": 0, "candidate_11": 0, "candidate_6": 0, "candidate_10": 0, "candidate_2": 0}, "ready": {"false": 0.9299999999999999, "true": 0.07}, "stage": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:ood:30", "state_id": "catalog_lookup_v2:ood:30", "family_id": "catalog_lookup_v2", "split": "ood", "state": "库存记录: item=SKU-00055; color=green; size=small; status=shipped; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "green, small", "candidate_1": "green, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_0", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 1, "candidate_1": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00055\",\"green\",\"small\",\"shipped\",\"no\"]", "template_id": "catalog_lookup_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_0": 1}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:ood:29", "state_id": "catalog_lookup_v2:ood:29", "family_id": "catalog_lookup_v2", "split": "ood", "state": "库存记录: item=SKU-00017; color=red; size=large; status=new; expedited=no.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "white, medium", "candidate_1": "yellow, medium", "candidate_2": "green, medium", "candidate_3": "cyan, medium", "candidate_4": "black, large", "candidate_5": "green, large", "candidate_6": "purple, large", "candidate_7": "blue, small", "candidate_8": "cyan, small", "candidate_9": "brown, small", "candidate_10": "gray, large", "candidate_11": "red, medium", "candidate_12": "yellow, small", "candidate_13": "pink, large", "candidate_14": "black, small", "candidate_15": "blue, large", "candidate_16": "blue, medium", "candidate_17": "red, large", "candidate_18": "cyan, large", "candidate_19": "brown, large"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_17", "ready": false, "stage": 0}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 0, "candidate_3": 0, "candidate_4": 0, "candidate_5": 0, "candidate_6": 0, "candidate_7": 0, "candidate_8": 0, "candidate_9": 0, "candidate_10": 0, "candidate_11": 0, "candidate_12": 0, "candidate_13": 0, "candidate_14": 0, "candidate_15": 0, "candidate_16": 0, "candidate_17": 1, "candidate_18": 0, "candidate_19": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00017\",\"red\",\"large\",\"new\",\"no\"]", "template_id": "catalog_lookup_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_14": 0, "candidate_8": 0, "candidate_2": 0, "candidate_7": 0, "candidate_6": 0, "candidate_15": 0, "candidate_13": 0, "candidate_12": 0, "candidate_9": 0, "candidate_3": 0, "candidate_18": 0, "candidate_17": 1, "candidate_4": 0, "candidate_19": 0, "candidate_11": 0, "candidate_0": 0, "candidate_16": 0, "candidate_10": 0, "candidate_5": 0}, "ready": {"false": 0.81, "true": 0.19}, "stage": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:ood:0", "state_id": "known_chance_v2:ood:0", "family_id": "known_chance_v2", "split": "ood", "state": "随机抽样实验: exactly one ball is drawn uniformly from a bag. Counts: blue=5, green=6, brown=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "brown": "The sampled ball is brown", "green": "The sampled ball is green"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue", "The drawn color is one of: green", "The drawn color is one of: brown"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.29411764705882354, "brown": 0.35294117647058826, "green": 0.35294117647058826}, "focus": {"false": 0.7058823529411764, "true": 0.29411764705882354}, "group": {"0": 0.29411764705882354, "1": 0.35294117647058826, "2": 0.35294117647058826}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"5/17\"],[\"brown\",\"6/17\"],[\"green\",\"6/17\"]]", "template_id": "known_chance_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"brown": 0.06, "blue": 0.59, "green": 0.35}, "focus": {"false": 0.75, "true": 0.25}, "group": {"0": 0.09, "1": 0.8, "2": 0.11}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:ood:1", "state_id": "known_chance_v2:ood:1", "family_id": "known_chance_v2", "split": "ood", "state": "随机抽样实验: exactly one ball is drawn uniformly from a bag. Counts: green=9, brown=7, pink=7, blue=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"brown": "The sampled ball is brown", "blue": "The sampled ball is blue", "pink": "The sampled ball is pink", "green": "The sampled ball is green"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a pink ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: green, blue", "The drawn color is one of: brown", "The drawn color is one of: pink"]}}, "gold": {}, "gold_probs": {"outcome": {"brown": 0.25925925925925924, "blue": 0.14814814814814814, "pink": 0.25925925925925924, "green": 0.3333333333333333}, "focus": {"false": 0.7407407407407407, "true": 0.25925925925925924}, "group": {"0": 0.48148148148148145, "1": 0.25925925925925924, "2": 0.25925925925925924}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"4/27\"],[\"brown\",\"7/27\"],[\"green\",\"1/3\"],[\"pink\",\"7/27\"]]", "template_id": "known_chance_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.01, "brown": 0.01, "blue": 0.01, "green": 0.97}, "focus": {"false": 0.79, "true": 0.21}, "group": {"0": 0.92, "1": 0.05, "2": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "catalog_lookup_v2:ood:31", "state_id": "catalog_lookup_v2:ood:31", "family_id": "catalog_lookup_v2", "split": "ood", "state": "库存记录: item=SKU-00086; color=yellow; size=medium; status=shipped; expedited=yes.", "questions": {"match": {"type": "choice", "instructions": "Choose the candidate whose color AND size both match this item. Ignore item code and shipping status.", "criteria": {"candidate_0": "red, small", "candidate_1": "black, small", "candidate_2": "yellow, medium", "candidate_3": "pink, large", "candidate_4": "black, medium"}}, "ready": {"type": "boolean", "instructions": "Is this item currently packed but not yet shipped? New, picked, and already shipped items do not qualify."}, "stage": {"type": "score", "instructions": "Select the stage explicitly reported in the record; expedited shipping does not change the current stage.", "criteria": ["New item, not picked", "Picked, not packed", "Packed, not shipped", "Already shipped"]}}, "gold": {"match": "candidate_2", "ready": false, "stage": 3}, "gold_probs": {"match": {"candidate_0": 0, "candidate_1": 0, "candidate_2": 1, "candidate_3": 0, "candidate_4": 0}, "ready": {"false": 1, "true": 0}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "gold_probs_kind": "deterministic_truth", "gold_label_kind": "deterministic_truth", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "lookup:[\"SKU-00086\",\"yellow\",\"medium\",\"shipped\",\"yes\"]", "template_id": "catalog_lookup_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"match": {"candidate_1": 0, "candidate_4": 0, "candidate_0": 0, "candidate_3": 0, "candidate_2": 1}, "ready": {"false": 0.98, "true": 0.02}, "stage": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:ood:4", "state_id": "known_chance_v2:ood:4", "family_id": "known_chance_v2", "split": "ood", "state": "随机抽样实验: exactly one ball is drawn uniformly from a bag. Counts: orange=2, green=8, cyan=8, yellow=6, black=8, pink=1, brown=1, blue=8, red=7, white=2, purple=3, gray=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"gray": "The sampled ball is gray", "orange": "The sampled ball is orange", "green": "The sampled ball is green", "black": "The sampled ball is black", "yellow": "The sampled ball is yellow", "white": "The sampled ball is white", "purple": "The sampled ball is purple", "red": "The sampled ball is red", "blue": "The sampled ball is blue", "brown": "The sampled ball is brown", "pink": "The sampled ball is pink", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a white ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: orange, yellow, brown, white", "The drawn color is one of: green, black, blue, purple", "The drawn color is one of: cyan, pink, red, gray"]}}, "gold": {}, "gold_probs": {"outcome": {"gray": 0.01818181818181818, "orange": 0.03636363636363636, "green": 0.14545454545454545, "black": 0.14545454545454545, "yellow": 0.10909090909090909, "white": 0.03636363636363636, "purple": 0.05454545454545454, "red": 0.12727272727272726, "blue": 0.14545454545454545, "brown": 0.01818181818181818, "pink": 0.01818181818181818, "cyan": 0.14545454545454545}, "focus": {"false": 0.9636363636363636, "true": 0.03636363636363636}, "group": {"0": 0.19999999999999998, "1": 0.4909090909090909, "2": 0.3090909090909091}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"8/55\"],[\"blue\",\"8/55\"],[\"brown\",\"1/55\"],[\"cyan\",\"8/55\"],[\"gray\",\"1/55\"],[\"green\",\"8/55\"],[\"orange\",\"2/55\"],[\"pink\",\"1/55\"],[\"purple\",\"3/55\"],[\"red\",\"7/55\"],[\"white\",\"2/55\"],[\"yellow\",\"6/55\"]]", "template_id": "known_chance_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0.14, "white": 0.01, "pink": 0.02, "yellow": 0.04, "green": 0.44, "orange": 0.03, "brown": 0.01, "purple": 0.01, "blue": 0.1, "gray": 0.01, "red": 0.12, "black": 0.07}, "focus": {"false": 0.91, "true": 0.09}, "group": {"0": 0.02, "1": 0.9, "2": 0.08}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:ood:3", "state_id": "known_chance_v2:ood:3", "family_id": "known_chance_v2", "split": "ood", "state": "随机抽样实验: exactly one ball is drawn uniformly from a bag. Counts: black=1, orange=2, blue=9, green=3, red=3, white=6, pink=1, brown=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "orange": "The sampled ball is orange", "blue": "The sampled ball is blue", "red": "The sampled ball is red", "brown": "The sampled ball is brown", "black": "The sampled ball is black", "pink": "The sampled ball is pink", "white": "The sampled ball is white"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a red ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: black, green, pink", "The drawn color is one of: orange, red, brown", "The drawn color is one of: blue, white"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.10344827586206896, "orange": 0.06896551724137931, "blue": 0.3103448275862069, "red": 0.10344827586206896, "brown": 0.13793103448275862, "black": 0.034482758620689655, "pink": 0.034482758620689655, "white": 0.20689655172413793}, "focus": {"false": 0.896551724137931, "true": 0.10344827586206896}, "group": {"0": 0.1724137931034483, "1": 0.3103448275862069, "2": 0.5172413793103449}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/29\"],[\"blue\",\"9/29\"],[\"brown\",\"4/29\"],[\"green\",\"3/29\"],[\"orange\",\"2/29\"],[\"pink\",\"1/29\"],[\"red\",\"3/29\"],[\"white\",\"6/29\"]]", "template_id": "known_chance_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"red": 0, "brown": 0, "blue": 0.95, "black": 0.03, "orange": 0, "white": 0.02, "green": 0, "pink": 0}, "focus": {"false": 0.85, "true": 0.15}, "group": {"0": 0, "1": 0.01, "2": 0.99}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:ood:5", "state_id": "known_chance_v2:ood:5", "family_id": "known_chance_v2", "split": "ood", "state": "随机抽样实验: exactly one ball is drawn uniformly from a bag. Counts: white=6, blue=4, yellow=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"white": "The sampled ball is white", "blue": "The sampled ball is blue", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white", "The drawn color is one of: blue", "The drawn color is one of: yellow"]}}, "gold": {}, "gold_probs": {"outcome": {"white": 0.5, "blue": 0.3333333333333333, "yellow": 0.16666666666666666}, "focus": {"false": 0.8333333333333334, "true": 0.16666666666666666}, "group": {"0": 0.5, "1": 0.3333333333333333, "2": 0.16666666666666666}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"1/3\"],[\"white\",\"1/2\"],[\"yellow\",\"1/6\"]]", "template_id": "known_chance_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"blue": 0.01, "white": 0.98, "yellow": 0.01}, "focus": {"false": 0.85, "true": 0.15}, "group": {"0": 0.96, "1": 0.02, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:ood:6", "state_id": "known_chance_v2:ood:6", "family_id": "known_chance_v2", "split": "ood", "state": "随机抽样实验: exactly one ball is drawn uniformly from a bag. Counts: cyan=6, brown=7, gray=9, pink=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "brown": "The sampled ball is brown", "gray": "The sampled ball is gray", "pink": "The sampled ball is pink"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: cyan, pink", "The drawn color is one of: brown", "The drawn color is one of: gray"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.24, "brown": 0.28, "gray": 0.36, "pink": 0.12}, "focus": {"false": 0.76, "true": 0.24}, "group": {"0": 0.36, "1": 0.28, "2": 0.36}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"7/25\"],[\"cyan\",\"6/25\"],[\"gray\",\"9/25\"],[\"pink\",\"3/25\"]]", "template_id": "known_chance_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.02, "brown": 0.04, "gray": 0.77, "cyan": 0.17}, "focus": {"false": 0.8200000000000001, "true": 0.18}, "group": {"0": 0.02, "1": 0.02, "2": 0.96}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:ood:8", "state_id": "known_chance_v2:ood:8", "family_id": "known_chance_v2", "split": "ood", "state": "随机抽样实验: exactly one ball is drawn uniformly from a bag. Counts: brown=7, gray=4, white=3, black=2, green=7, red=3, orange=4, blue=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "green": "The sampled ball is green", "orange": "The sampled ball is orange", "gray": "The sampled ball is gray", "white": "The sampled ball is white", "brown": "The sampled ball is brown", "red": "The sampled ball is red", "blue": "The sampled ball is blue"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a green ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: brown, black, orange", "The drawn color is one of: gray, green, blue", "The drawn color is one of: white, red"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.05555555555555555, "green": 0.19444444444444445, "orange": 0.1111111111111111, "gray": 0.1111111111111111, "white": 0.08333333333333333, "brown": 0.19444444444444445, "red": 0.08333333333333333, "blue": 0.16666666666666666}, "focus": {"false": 0.8055555555555556, "true": 0.19444444444444445}, "group": {"0": 0.3611111111111111, "1": 0.4722222222222222, "2": 0.16666666666666666}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/18\"],[\"blue\",\"1/6\"],[\"brown\",\"7/36\"],[\"gray\",\"1/9\"],[\"green\",\"7/36\"],[\"orange\",\"1/9\"],[\"red\",\"1/12\"],[\"white\",\"1/12\"]]", "template_id": "known_chance_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0, "green": 0.15, "gray": 0.01, "red": 0, "orange": 0, "brown": 0.73, "blue": 0.09, "black": 0.02}, "focus": {"false": 0.81, "true": 0.19}, "group": {"0": 0.11, "1": 0.88, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:ood:7", "state_id": "known_chance_v2:ood:7", "family_id": "known_chance_v2", "split": "ood", "state": "随机抽样实验: exactly one ball is drawn uniformly from a bag. Counts: white=6, blue=6, black=1, cyan=6, brown=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "cyan": "The sampled ball is cyan", "blue": "The sampled ball is blue", "white": "The sampled ball is white", "brown": "The sampled ball is brown"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white, cyan", "The drawn color is one of: blue, brown", "The drawn color is one of: black"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.041666666666666664, "cyan": 0.25, "blue": 0.25, "white": 0.25, "brown": 0.20833333333333334}, "focus": {"false": 0.9583333333333334, "true": 0.041666666666666664}, "group": {"0": 0.5, "1": 0.45833333333333337, "2": 0.041666666666666664}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/24\"],[\"blue\",\"1/4\"],[\"brown\",\"5/24\"],[\"cyan\",\"1/4\"],[\"white\",\"1/4\"]]", "template_id": "known_chance_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0.03, "brown": 0.04, "blue": 0.07, "black": 0.08, "white": 0.78}, "focus": {"false": 0.9299999999999999, "true": 0.07}, "group": {"0": 0.92, "1": 0.05, "2": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:ood:2", "state_id": "known_chance_v2:ood:2", "family_id": "known_chance_v2", "split": "ood", "state": "随机抽样实验: exactly one ball is drawn uniformly from a bag. Counts: blue=1, white=2, green=3, red=4, brown=2. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "blue": "The sampled ball is blue", "white": "The sampled ball is white", "brown": "The sampled ball is brown", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a brown ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: blue, red", "The drawn color is one of: white, brown", "The drawn color is one of: green"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.25, "blue": 0.08333333333333333, "white": 0.16666666666666666, "brown": 0.16666666666666666, "red": 0.3333333333333333}, "focus": {"false": 0.8333333333333334, "true": 0.16666666666666666}, "group": {"0": 0.41666666666666663, "1": 0.3333333333333333, "2": 0.25}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"1/12\"],[\"brown\",\"1/6\"],[\"green\",\"1/4\"],[\"red\",\"1/3\"],[\"white\",\"1/6\"]]", "template_id": "known_chance_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"white": 0, "brown": 0.01, "blue": 0.2, "green": 0.02, "red": 0.77}, "focus": {"false": 0.81, "true": 0.19}, "group": {"0": 0.78, "1": 0.11, "2": 0.11}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:ood:10", "state_id": "known_chance_v2:ood:10", "family_id": "known_chance_v2", "split": "ood", "state": "随机抽样实验: exactly one ball is drawn uniformly from a bag. Counts: pink=1, purple=7, red=1. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"purple": "The sampled ball is purple", "pink": "The sampled ball is pink", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a pink ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: pink", "The drawn color is one of: purple", "The drawn color is one of: red"]}}, "gold": {}, "gold_probs": {"outcome": {"purple": 0.7777777777777778, "pink": 0.1111111111111111, "red": 0.1111111111111111}, "focus": {"false": 0.8888888888888888, "true": 0.1111111111111111}, "group": {"0": 0.1111111111111111, "1": 0.7777777777777778, "2": 0.1111111111111111}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"pink\",\"1/9\"],[\"purple\",\"7/9\"],[\"red\",\"1/9\"]]", "template_id": "known_chance_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.01, "red": 0, "purple": 0.99}, "focus": {"false": 0.91, "true": 0.09}, "group": {"0": 0, "1": 1, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:ood:9", "state_id": "known_chance_v2:ood:9", "family_id": "known_chance_v2", "split": "ood", "state": "随机抽样实验: exactly one ball is drawn uniformly from a bag. Counts: red=9, brown=7, pink=9, orange=4, purple=3, cyan=9, black=7, yellow=5, gray=1, white=6, green=6, blue=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"orange": "The sampled ball is orange", "pink": "The sampled ball is pink", "brown": "The sampled ball is brown", "blue": "The sampled ball is blue", "black": "The sampled ball is black", "yellow": "The sampled ball is yellow", "green": "The sampled ball is green", "red": "The sampled ball is red", "gray": "The sampled ball is gray", "cyan": "The sampled ball is cyan", "purple": "The sampled ball is purple", "white": "The sampled ball is white"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: red, orange, black, white", "The drawn color is one of: brown, purple, yellow, green", "The drawn color is one of: pink, cyan, gray, blue"]}}, "gold": {}, "gold_probs": {"outcome": {"orange": 0.05555555555555555, "pink": 0.125, "brown": 0.09722222222222222, "blue": 0.08333333333333333, "black": 0.09722222222222222, "yellow": 0.06944444444444445, "green": 0.08333333333333333, "red": 0.125, "gray": 0.013888888888888888, "cyan": 0.125, "purple": 0.041666666666666664, "white": 0.08333333333333333}, "focus": {"false": 0.9305555555555556, "true": 0.06944444444444445}, "group": {"0": 0.3611111111111111, "1": 0.2916666666666667, "2": 0.3472222222222222}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"7/72\"],[\"blue\",\"1/12\"],[\"brown\",\"7/72\"],[\"cyan\",\"1/8\"],[\"gray\",\"1/72\"],[\"green\",\"1/12\"],[\"orange\",\"1/18\"],[\"pink\",\"1/8\"],[\"purple\",\"1/24\"],[\"red\",\"1/8\"],[\"white\",\"1/12\"],[\"yellow\",\"5/72\"]]", "template_id": "known_chance_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"gray": 0.21, "purple": 0.01, "yellow": 0, "black": 0.01, "cyan": 0.04, "white": 0, "blue": 0.01, "orange": 0.01, "green": 0.01, "red": 0.64, "pink": 0.05, "brown": 0.01}, "focus": {"false": 0.89, "true": 0.11}, "group": {"0": 0.55, "1": 0.06, "2": 0.39}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:ood:12", "state_id": "known_chance_v2:ood:12", "family_id": "known_chance_v2", "split": "ood", "state": "随机抽样实验: exactly one ball is drawn uniformly from a bag. Counts: green=8, yellow=5, pink=3, cyan=4, blue=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"pink": "The sampled ball is pink", "cyan": "The sampled ball is cyan", "blue": "The sampled ball is blue", "green": "The sampled ball is green", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a green ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: green, cyan", "The drawn color is one of: yellow, blue", "The drawn color is one of: pink"]}}, "gold": {}, "gold_probs": {"outcome": {"pink": 0.125, "cyan": 0.16666666666666666, "blue": 0.16666666666666666, "green": 0.3333333333333333, "yellow": 0.20833333333333334}, "focus": {"false": 0.6666666666666667, "true": 0.3333333333333333}, "group": {"0": 0.5, "1": 0.375, "2": 0.125}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"1/6\"],[\"cyan\",\"1/6\"],[\"green\",\"1/3\"],[\"pink\",\"1/8\"],[\"yellow\",\"5/24\"]]", "template_id": "known_chance_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"blue": 0, "green": 1, "cyan": 0, "pink": 0, "yellow": 0}, "focus": {"false": 0.72, "true": 0.28}, "group": {"0": 0.96, "1": 0.02, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:ood:13", "state_id": "known_chance_v2:ood:13", "family_id": "known_chance_v2", "split": "ood", "state": "随机抽样实验: exactly one ball is drawn uniformly from a bag. Counts: cyan=4, white=2, orange=8, yellow=3, red=1, green=7, blue=8, black=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "blue": "The sampled ball is blue", "yellow": "The sampled ball is yellow", "white": "The sampled ball is white", "black": "The sampled ball is black", "green": "The sampled ball is green", "orange": "The sampled ball is orange", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a orange ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: cyan, yellow, blue", "The drawn color is one of: white, red, black", "The drawn color is one of: orange, green"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.10810810810810811, "blue": 0.21621621621621623, "yellow": 0.08108108108108109, "white": 0.05405405405405406, "black": 0.10810810810810811, "green": 0.1891891891891892, "orange": 0.21621621621621623, "red": 0.02702702702702703}, "focus": {"false": 0.7837837837837838, "true": 0.21621621621621623}, "group": {"0": 0.40540540540540543, "1": 0.1891891891891892, "2": 0.40540540540540543}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"4/37\"],[\"blue\",\"8/37\"],[\"cyan\",\"4/37\"],[\"green\",\"7/37\"],[\"orange\",\"8/37\"],[\"red\",\"1/37\"],[\"white\",\"2/37\"],[\"yellow\",\"3/37\"]]", "template_id": "known_chance_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0.03, "yellow": 0.01, "black": 0.01, "orange": 0.59, "cyan": 0.05, "blue": 0.2, "white": 0.02, "red": 0.09}, "focus": {"false": 0.79, "true": 0.21}, "group": {"0": 0.23, "1": 0.02, "2": 0.75}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:ood:11", "state_id": "known_chance_v2:ood:11", "family_id": "known_chance_v2", "split": "ood", "state": "随机抽样实验: exactly one ball is drawn uniformly from a bag. Counts: yellow=6, purple=6, orange=2, brown=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"orange": "The sampled ball is orange", "purple": "The sampled ball is purple", "brown": "The sampled ball is brown", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a orange ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: yellow, brown", "The drawn color is one of: purple", "The drawn color is one of: orange"]}}, "gold": {}, "gold_probs": {"outcome": {"orange": 0.1, "purple": 0.3, "brown": 0.3, "yellow": 0.3}, "focus": {"false": 0.9, "true": 0.1}, "group": {"0": 0.6, "1": 0.3, "2": 0.1}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"3/10\"],[\"orange\",\"1/10\"],[\"purple\",\"3/10\"],[\"yellow\",\"3/10\"]]", "template_id": "known_chance_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"yellow": 0.76, "brown": 0.03, "purple": 0.11, "orange": 0.1}, "focus": {"false": 0.87, "true": 0.13}, "group": {"0": 0.94, "1": 0.03, "2": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:ood:14", "state_id": "known_chance_v2:ood:14", "family_id": "known_chance_v2", "split": "ood", "state": "随机抽样实验: exactly one ball is drawn uniformly from a bag. Counts: black=7, yellow=3, brown=7, cyan=9, purple=6, white=3, green=9, orange=4, gray=6, blue=9, pink=7, red=3. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"orange": "The sampled ball is orange", "brown": "The sampled ball is brown", "black": "The sampled ball is black", "green": "The sampled ball is green", "purple": "The sampled ball is purple", "red": "The sampled ball is red", "gray": "The sampled ball is gray", "yellow": "The sampled ball is yellow", "pink": "The sampled ball is pink", "blue": "The sampled ball is blue", "cyan": "The sampled ball is cyan", "white": "The sampled ball is white"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a gray ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: black, cyan, green, blue", "The drawn color is one of: yellow, purple, orange, pink", "The drawn color is one of: brown, white, gray, red"]}}, "gold": {}, "gold_probs": {"outcome": {"orange": 0.0547945205479452, "brown": 0.0958904109589041, "black": 0.0958904109589041, "green": 0.1232876712328767, "purple": 0.0821917808219178, "red": 0.0410958904109589, "gray": 0.0821917808219178, "yellow": 0.0410958904109589, "pink": 0.0958904109589041, "blue": 0.1232876712328767, "cyan": 0.1232876712328767, "white": 0.0410958904109589}, "focus": {"false": 0.9178082191780822, "true": 0.0821917808219178}, "group": {"0": 0.4657534246575342, "1": 0.273972602739726, "2": 0.2602739726027397}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"7/73\"],[\"blue\",\"9/73\"],[\"brown\",\"7/73\"],[\"cyan\",\"9/73\"],[\"gray\",\"6/73\"],[\"green\",\"9/73\"],[\"orange\",\"4/73\"],[\"pink\",\"7/73\"],[\"purple\",\"6/73\"],[\"red\",\"3/73\"],[\"white\",\"3/73\"],[\"yellow\",\"3/73\"]]", "template_id": "known_chance_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"black": 0.4, "red": 0.01, "gray": 0.01, "orange": 0.01, "green": 0.1, "white": 0.01, "blue": 0.08, "yellow": 0.01, "brown": 0.02, "purple": 0.01, "pink": 0.01, "cyan": 0.33}, "focus": {"false": 0.88, "true": 0.12}, "group": {"0": 0.98, "1": 0.01, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:ood:15", "state_id": "known_chance_v2:ood:15", "family_id": "known_chance_v2", "split": "ood", "state": "随机抽样实验: exactly one ball is drawn uniformly from a bag. Counts: red=3, white=9, blue=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "white": "The sampled ball is white", "red": "The sampled ball is red"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a blue ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: red", "The drawn color is one of: white", "The drawn color is one of: blue"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.29411764705882354, "white": 0.5294117647058824, "red": 0.17647058823529413}, "focus": {"false": 0.7058823529411764, "true": 0.29411764705882354}, "group": {"0": 0.17647058823529413, "1": 0.5294117647058824, "2": 0.29411764705882354}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"5/17\"],[\"red\",\"3/17\"],[\"white\",\"9/17\"]]", "template_id": "known_chance_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"blue": 0.01, "white": 0.89, "red": 0.1}, "focus": {"false": 0.78, "true": 0.22}, "group": {"0": 0.02, "1": 0.97, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:ood:17", "state_id": "known_chance_v2:ood:17", "family_id": "known_chance_v2", "split": "ood", "state": "随机抽样实验: exactly one ball is drawn uniformly from a bag. Counts: brown=2, gray=2, blue=5, orange=6, white=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"orange": "The sampled ball is orange", "white": "The sampled ball is white", "brown": "The sampled ball is brown", "blue": "The sampled ball is blue", "gray": "The sampled ball is gray"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a gray ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: brown, orange", "The drawn color is one of: gray, white", "The drawn color is one of: blue"]}}, "gold": {}, "gold_probs": {"outcome": {"orange": 0.3157894736842105, "white": 0.21052631578947367, "brown": 0.10526315789473684, "blue": 0.2631578947368421, "gray": 0.10526315789473684}, "focus": {"false": 0.8947368421052632, "true": 0.10526315789473684}, "group": {"0": 0.42105263157894735, "1": 0.3157894736842105, "2": 0.2631578947368421}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"5/19\"],[\"brown\",\"2/19\"],[\"gray\",\"2/19\"],[\"orange\",\"6/19\"],[\"white\",\"4/19\"]]", "template_id": "known_chance_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"orange": 0.86, "blue": 0.07, "white": 0.01, "brown": 0.06, "gray": 0}, "focus": {"false": 0.85, "true": 0.15}, "group": {"0": 0.67, "1": 0.05, "2": 0.28}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:ood:18", "state_id": "known_chance_v2:ood:18", "family_id": "known_chance_v2", "split": "ood", "state": "随机抽样实验: exactly one ball is drawn uniformly from a bag. Counts: yellow=4, white=2, green=4, black=7, pink=1, gray=2, red=7, cyan=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"pink": "The sampled ball is pink", "white": "The sampled ball is white", "cyan": "The sampled ball is cyan", "gray": "The sampled ball is gray", "black": "The sampled ball is black", "red": "The sampled ball is red", "green": "The sampled ball is green", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: yellow, black, red", "The drawn color is one of: white, pink, cyan", "The drawn color is one of: green, gray"]}}, "gold": {}, "gold_probs": {"outcome": {"pink": 0.030303030303030304, "white": 0.06060606060606061, "cyan": 0.18181818181818182, "gray": 0.06060606060606061, "black": 0.21212121212121213, "red": 0.21212121212121213, "green": 0.12121212121212122, "yellow": 0.12121212121212122}, "focus": {"false": 0.8787878787878788, "true": 0.12121212121212122}, "group": {"0": 0.5454545454545454, "1": 0.2727272727272727, "2": 0.18181818181818182}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"7/33\"],[\"cyan\",\"2/11\"],[\"gray\",\"2/33\"],[\"green\",\"4/33\"],[\"pink\",\"1/33\"],[\"red\",\"7/33\"],[\"white\",\"2/33\"],[\"yellow\",\"4/33\"]]", "template_id": "known_chance_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.01, "green": 0.01, "cyan": 0.02, "red": 0.33, "white": 0, "black": 0.54, "yellow": 0.09, "gray": 0}, "focus": {"false": 0.85, "true": 0.15}, "group": {"0": 0.99, "1": 0.01, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:ood:20", "state_id": "known_chance_v2:ood:20", "family_id": "known_chance_v2", "split": "ood", "state": "随机抽样实验: exactly one ball is drawn uniformly from a bag. Counts: pink=2, brown=8, red=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"brown": "The sampled ball is brown", "red": "The sampled ball is red", "pink": "The sampled ball is pink"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a brown ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: pink", "The drawn color is one of: brown", "The drawn color is one of: red"]}}, "gold": {}, "gold_probs": {"outcome": {"brown": 0.47058823529411764, "red": 0.4117647058823529, "pink": 0.11764705882352941}, "focus": {"false": 0.5294117647058824, "true": 0.47058823529411764}, "group": {"0": 0.11764705882352941, "1": 0.47058823529411764, "2": 0.4117647058823529}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"brown\",\"8/17\"],[\"pink\",\"2/17\"],[\"red\",\"7/17\"]]", "template_id": "known_chance_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"red": 0.03, "brown": 0.93, "pink": 0.04}, "focus": {"false": 0.6, "true": 0.4}, "group": {"0": 0.01, "1": 0.96, "2": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:ood:19", "state_id": "known_chance_v2:ood:19", "family_id": "known_chance_v2", "split": "ood", "state": "随机抽样实验: exactly one ball is drawn uniformly from a bag. Counts: yellow=6, cyan=6, white=8, brown=8, purple=1, gray=7, black=6, pink=1, orange=5, blue=8, green=1, red=9. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"green": "The sampled ball is green", "red": "The sampled ball is red", "blue": "The sampled ball is blue", "purple": "The sampled ball is purple", "yellow": "The sampled ball is yellow", "gray": "The sampled ball is gray", "black": "The sampled ball is black", "white": "The sampled ball is white", "cyan": "The sampled ball is cyan", "brown": "The sampled ball is brown", "pink": "The sampled ball is pink", "orange": "The sampled ball is orange"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a white ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: yellow, brown, black, blue", "The drawn color is one of: cyan, purple, pink, green", "The drawn color is one of: white, gray, orange, red"]}}, "gold": {}, "gold_probs": {"outcome": {"green": 0.015151515151515152, "red": 0.13636363636363635, "blue": 0.12121212121212122, "purple": 0.015151515151515152, "yellow": 0.09090909090909091, "gray": 0.10606060606060606, "black": 0.09090909090909091, "white": 0.12121212121212122, "cyan": 0.09090909090909091, "brown": 0.12121212121212122, "pink": 0.015151515151515152, "orange": 0.07575757575757576}, "focus": {"false": 0.8787878787878788, "true": 0.12121212121212122}, "group": {"0": 0.42424242424242425, "1": 0.13636363636363635, "2": 0.4393939393939394}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/11\"],[\"blue\",\"4/33\"],[\"brown\",\"4/33\"],[\"cyan\",\"1/11\"],[\"gray\",\"7/66\"],[\"green\",\"1/66\"],[\"orange\",\"5/66\"],[\"pink\",\"1/66\"],[\"purple\",\"1/66\"],[\"red\",\"3/22\"],[\"white\",\"4/33\"],[\"yellow\",\"1/11\"]]", "template_id": "known_chance_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"red": 0.8, "brown": 0.01, "blue": 0.01, "gray": 0.02, "black": 0, "yellow": 0.09, "orange": 0.01, "purple": 0.01, "cyan": 0.01, "white": 0.03, "pink": 0, "green": 0.01}, "focus": {"false": 0.85, "true": 0.15}, "group": {"0": 0.21, "1": 0.01, "2": 0.78}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:ood:21", "state_id": "known_chance_v2:ood:21", "family_id": "known_chance_v2", "split": "ood", "state": "随机抽样实验: exactly one ball is drawn uniformly from a bag. Counts: purple=2, black=2, orange=3, yellow=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "orange": "The sampled ball is orange", "yellow": "The sampled ball is yellow", "purple": "The sampled ball is purple"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: purple, yellow", "The drawn color is one of: black", "The drawn color is one of: orange"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.18181818181818182, "orange": 0.2727272727272727, "yellow": 0.36363636363636365, "purple": 0.18181818181818182}, "focus": {"false": 0.6363636363636364, "true": 0.36363636363636365}, "group": {"0": 0.5454545454545454, "1": 0.18181818181818182, "2": 0.2727272727272727}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"2/11\"],[\"orange\",\"3/11\"],[\"purple\",\"2/11\"],[\"yellow\",\"4/11\"]]", "template_id": "known_chance_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"orange": 0.05, "purple": 0.15, "black": 0.03, "yellow": 0.77}, "focus": {"false": 0.73, "true": 0.27}, "group": {"0": 0.74, "1": 0.01, "2": 0.25}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:ood:23", "state_id": "known_chance_v2:ood:23", "family_id": "known_chance_v2", "split": "ood", "state": "随机抽样实验: exactly one ball is drawn uniformly from a bag. Counts: white=5, yellow=2, cyan=7, purple=7, blue=5, black=9, brown=5, pink=9. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"brown": "The sampled ball is brown", "cyan": "The sampled ball is cyan", "pink": "The sampled ball is pink", "white": "The sampled ball is white", "black": "The sampled ball is black", "blue": "The sampled ball is blue", "yellow": "The sampled ball is yellow", "purple": "The sampled ball is purple"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a brown ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white, purple, brown", "The drawn color is one of: yellow, blue, pink", "The drawn color is one of: cyan, black"]}}, "gold": {}, "gold_probs": {"outcome": {"brown": 0.10204081632653061, "cyan": 0.14285714285714285, "pink": 0.1836734693877551, "white": 0.10204081632653061, "black": 0.1836734693877551, "blue": 0.10204081632653061, "yellow": 0.04081632653061224, "purple": 0.14285714285714285}, "focus": {"false": 0.8979591836734694, "true": 0.10204081632653061}, "group": {"0": 0.3469387755102041, "1": 0.326530612244898, "2": 0.326530612244898}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"9/49\"],[\"blue\",\"5/49\"],[\"brown\",\"5/49\"],[\"cyan\",\"1/7\"],[\"pink\",\"9/49\"],[\"purple\",\"1/7\"],[\"white\",\"5/49\"],[\"yellow\",\"2/49\"]]", "template_id": "known_chance_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0.07, "blue": 0.01, "purple": 0.02, "pink": 0.11, "white": 0.2, "brown": 0.02, "black": 0.47, "yellow": 0.1}, "focus": {"false": 0.87, "true": 0.13}, "group": {"0": 0.05, "1": 0.05, "2": 0.9}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:ood:22", "state_id": "known_chance_v2:ood:22", "family_id": "known_chance_v2", "split": "ood", "state": "随机抽样实验: exactly one ball is drawn uniformly from a bag. Counts: brown=1, yellow=9, orange=1, red=5, blue=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"red": "The sampled ball is red", "blue": "The sampled ball is blue", "yellow": "The sampled ball is yellow", "brown": "The sampled ball is brown", "orange": "The sampled ball is orange"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: brown, red", "The drawn color is one of: yellow, blue", "The drawn color is one of: orange"]}}, "gold": {}, "gold_probs": {"outcome": {"red": 0.25, "blue": 0.2, "yellow": 0.45, "brown": 0.05, "orange": 0.05}, "focus": {"false": 0.55, "true": 0.45}, "group": {"0": 0.3, "1": 0.65, "2": 0.05}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"blue\",\"1/5\"],[\"brown\",\"1/20\"],[\"orange\",\"1/20\"],[\"red\",\"1/4\"],[\"yellow\",\"9/20\"]]", "template_id": "known_chance_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"orange": 0, "yellow": 0.97, "red": 0.01, "blue": 0, "brown": 0.02}, "focus": {"false": 0.5700000000000001, "true": 0.43}, "group": {"0": 0.02, "1": 0.98, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:ood:16", "state_id": "known_chance_v2:ood:16", "family_id": "known_chance_v2", "split": "ood", "state": "随机抽样实验: exactly one ball is drawn uniformly from a bag. Counts: white=2, cyan=3, purple=6, red=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"cyan": "The sampled ball is cyan", "white": "The sampled ball is white", "red": "The sampled ball is red", "purple": "The sampled ball is purple"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a white ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white, red", "The drawn color is one of: cyan", "The drawn color is one of: purple"]}}, "gold": {}, "gold_probs": {"outcome": {"cyan": 0.2, "white": 0.13333333333333333, "red": 0.26666666666666666, "purple": 0.4}, "focus": {"false": 0.8666666666666667, "true": 0.13333333333333333}, "group": {"0": 0.4, "1": 0.2, "2": 0.4}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"1/5\"],[\"purple\",\"2/5\"],[\"red\",\"4/15\"],[\"white\",\"2/15\"]]", "template_id": "known_chance_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.83, "red": 0.01, "cyan": 0.02, "white": 0.14}, "focus": {"false": 0.86, "true": 0.14}, "group": {"0": 0.01, "1": 0.01, "2": 0.98}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:ood:25", "state_id": "known_chance_v2:ood:25", "family_id": "known_chance_v2", "split": "ood", "state": "随机抽样实验: exactly one ball is drawn uniformly from a bag. Counts: purple=4, cyan=7, yellow=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"purple": "The sampled ball is purple", "cyan": "The sampled ball is cyan", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a purple ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: purple", "The drawn color is one of: cyan", "The drawn color is one of: yellow"]}}, "gold": {}, "gold_probs": {"outcome": {"purple": 0.25, "cyan": 0.4375, "yellow": 0.3125}, "focus": {"false": 0.75, "true": 0.25}, "group": {"0": 0.25, "1": 0.4375, "2": 0.3125}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"7/16\"],[\"purple\",\"1/4\"],[\"yellow\",\"5/16\"]]", "template_id": "known_chance_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.2, "yellow": 0.01, "cyan": 0.79}, "focus": {"false": 0.81, "true": 0.19}, "group": {"0": 0.03, "1": 0.95, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:ood:24", "state_id": "known_chance_v2:ood:24", "family_id": "known_chance_v2", "split": "ood", "state": "随机抽样实验: exactly one ball is drawn uniformly from a bag. Counts: gray=4, white=8, cyan=7, orange=9, yellow=6, purple=7, blue=2, red=7, pink=5, black=8, green=4, brown=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"orange": "The sampled ball is orange", "pink": "The sampled ball is pink", "red": "The sampled ball is red", "gray": "The sampled ball is gray", "yellow": "The sampled ball is yellow", "blue": "The sampled ball is blue", "white": "The sampled ball is white", "black": "The sampled ball is black", "purple": "The sampled ball is purple", "green": "The sampled ball is green", "brown": "The sampled ball is brown", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a orange ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: gray, orange, blue, black", "The drawn color is one of: white, yellow, red, green", "The drawn color is one of: cyan, purple, pink, brown"]}}, "gold": {}, "gold_probs": {"outcome": {"orange": 0.1267605633802817, "pink": 0.07042253521126761, "red": 0.09859154929577464, "gray": 0.056338028169014086, "yellow": 0.08450704225352113, "blue": 0.028169014084507043, "white": 0.11267605633802817, "black": 0.11267605633802817, "purple": 0.09859154929577464, "green": 0.056338028169014086, "brown": 0.056338028169014086, "cyan": 0.09859154929577464}, "focus": {"false": 0.8732394366197183, "true": 0.1267605633802817}, "group": {"0": 0.323943661971831, "1": 0.352112676056338, "2": 0.323943661971831}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"8/71\"],[\"blue\",\"2/71\"],[\"brown\",\"4/71\"],[\"cyan\",\"7/71\"],[\"gray\",\"4/71\"],[\"green\",\"4/71\"],[\"orange\",\"9/71\"],[\"pink\",\"5/71\"],[\"purple\",\"7/71\"],[\"red\",\"7/71\"],[\"white\",\"8/71\"],[\"yellow\",\"6/71\"]]", "template_id": "known_chance_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"blue": 0.06, "red": 0.01, "gray": 0.07, "orange": 0.74, "green": 0.01, "white": 0.06, "black": 0.02, "yellow": 0, "cyan": 0.01, "purple": 0.01, "brown": 0.01, "pink": 0}, "focus": {"false": 0.85, "true": 0.15}, "group": {"0": 0.37, "1": 0.41, "2": 0.22}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:ood:26", "state_id": "known_chance_v2:ood:26", "family_id": "known_chance_v2", "split": "ood", "state": "随机抽样实验: exactly one ball is drawn uniformly from a bag. Counts: pink=2, blue=8, black=9, yellow=5. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "yellow": "The sampled ball is yellow", "pink": "The sampled ball is pink", "black": "The sampled ball is black"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a yellow ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: pink, yellow", "The drawn color is one of: blue", "The drawn color is one of: black"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.3333333333333333, "yellow": 0.20833333333333334, "pink": 0.08333333333333333, "black": 0.375}, "focus": {"false": 0.7916666666666666, "true": 0.20833333333333334}, "group": {"0": 0.2916666666666667, "1": 0.3333333333333333, "2": 0.375}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"3/8\"],[\"blue\",\"1/3\"],[\"pink\",\"1/12\"],[\"yellow\",\"5/24\"]]", "template_id": "known_chance_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.07, "black": 0.84, "blue": 0.08, "yellow": 0.01}, "focus": {"false": 0.84, "true": 0.16}, "group": {"0": 0.01, "1": 0.04, "2": 0.95}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:ood:27", "state_id": "known_chance_v2:ood:27", "family_id": "known_chance_v2", "split": "ood", "state": "随机抽样实验: exactly one ball is drawn uniformly from a bag. Counts: black=9, blue=3, gray=1, purple=6, red=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"blue": "The sampled ball is blue", "red": "The sampled ball is red", "black": "The sampled ball is black", "gray": "The sampled ball is gray", "purple": "The sampled ball is purple"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a gray ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: black, purple", "The drawn color is one of: blue, red", "The drawn color is one of: gray"]}}, "gold": {}, "gold_probs": {"outcome": {"blue": 0.12, "red": 0.24, "black": 0.36, "gray": 0.04, "purple": 0.24}, "focus": {"false": 0.96, "true": 0.04}, "group": {"0": 0.6, "1": 0.36, "2": 0.04}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"9/25\"],[\"blue\",\"3/25\"],[\"gray\",\"1/25\"],[\"purple\",\"6/25\"],[\"red\",\"6/25\"]]", "template_id": "known_chance_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"gray": 0.03, "purple": 0.01, "blue": 0, "red": 0.02, "black": 0.93}, "focus": {"false": 0.9299999999999999, "true": 0.07}, "group": {"0": 0.98, "1": 0.01, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:ood:29", "state_id": "known_chance_v2:ood:29", "family_id": "known_chance_v2", "split": "ood", "state": "随机抽样实验: exactly one ball is drawn uniformly from a bag. Counts: purple=9, white=1, gray=7, red=2, black=9, yellow=3, blue=6, cyan=7, brown=9, orange=9, pink=6, green=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "white": "The sampled ball is white", "green": "The sampled ball is green", "yellow": "The sampled ball is yellow", "pink": "The sampled ball is pink", "gray": "The sampled ball is gray", "purple": "The sampled ball is purple", "red": "The sampled ball is red", "blue": "The sampled ball is blue", "cyan": "The sampled ball is cyan", "brown": "The sampled ball is brown", "orange": "The sampled ball is orange"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a green ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: purple, red, blue, orange", "The drawn color is one of: white, black, cyan, pink", "The drawn color is one of: gray, yellow, brown, green"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.12, "white": 0.013333333333333334, "green": 0.09333333333333334, "yellow": 0.04, "pink": 0.08, "gray": 0.09333333333333334, "purple": 0.12, "red": 0.02666666666666667, "blue": 0.08, "cyan": 0.09333333333333334, "brown": 0.12, "orange": 0.12}, "focus": {"false": 0.9066666666666666, "true": 0.09333333333333334}, "group": {"0": 0.3466666666666667, "1": 0.30666666666666664, "2": 0.3466666666666667}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"3/25\"],[\"blue\",\"2/25\"],[\"brown\",\"3/25\"],[\"cyan\",\"7/75\"],[\"gray\",\"7/75\"],[\"green\",\"7/75\"],[\"orange\",\"3/25\"],[\"pink\",\"2/25\"],[\"purple\",\"3/25\"],[\"red\",\"2/75\"],[\"white\",\"1/75\"],[\"yellow\",\"1/25\"]]", "template_id": "known_chance_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"purple": 0.65, "cyan": 0.01, "white": 0.08, "pink": 0.01, "gray": 0.02, "yellow": 0.01, "brown": 0.08, "blue": 0.01, "red": 0.02, "orange": 0.03, "green": 0.02, "black": 0.06}, "focus": {"false": 0.87, "true": 0.13}, "group": {"0": 0.48, "1": 0.1, "2": 0.42}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:ood:30", "state_id": "known_chance_v2:ood:30", "family_id": "known_chance_v2", "split": "ood", "state": "随机抽样实验: exactly one ball is drawn uniformly from a bag. Counts: yellow=2, pink=3, black=4. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "pink": "The sampled ball is pink", "yellow": "The sampled ball is yellow"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a black ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: yellow", "The drawn color is one of: pink", "The drawn color is one of: black"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.4444444444444444, "pink": 0.3333333333333333, "yellow": 0.2222222222222222}, "focus": {"false": 0.5555555555555556, "true": 0.4444444444444444}, "group": {"0": 0.2222222222222222, "1": 0.3333333333333333, "2": 0.4444444444444444}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"4/9\"],[\"pink\",\"1/3\"],[\"yellow\",\"2/9\"]]", "template_id": "known_chance_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"pink": 0.06, "yellow": 0.19, "black": 0.75}, "focus": {"false": 0.6599999999999999, "true": 0.34}, "group": {"0": 0.04, "1": 0.03, "2": 0.93}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:ood:31", "state_id": "known_chance_v2:ood:31", "family_id": "known_chance_v2", "split": "ood", "state": "随机抽样实验: exactly one ball is drawn uniformly from a bag. Counts: white=6, black=1, green=6, cyan=7. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"black": "The sampled ball is black", "white": "The sampled ball is white", "green": "The sampled ball is green", "cyan": "The sampled ball is cyan"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a cyan ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: white, cyan", "The drawn color is one of: black", "The drawn color is one of: green"]}}, "gold": {}, "gold_probs": {"outcome": {"black": 0.05, "white": 0.3, "green": 0.3, "cyan": 0.35}, "focus": {"false": 0.65, "true": 0.35}, "group": {"0": 0.6499999999999999, "1": 0.05, "2": 0.3}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"black\",\"1/20\"],[\"cyan\",\"7/20\"],[\"green\",\"3/10\"],[\"white\",\"3/10\"]]", "template_id": "known_chance_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"cyan": 0.69, "green": 0.02, "black": 0.05, "white": 0.24}, "focus": {"false": 0.74, "true": 0.26}, "group": {"0": 0.97, "1": 0.01, "2": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "known_chance_v2:ood:28", "state_id": "known_chance_v2:ood:28", "family_id": "known_chance_v2", "split": "ood", "state": "随机抽样实验: exactly one ball is drawn uniformly from a bag. Counts: cyan=1, green=5, orange=9, pink=7, yellow=8, white=5, red=7, purple=6. All possible colors are listed; the draw has NOT happened.", "questions": {"outcome": {"type": "choice", "instructions": "Predict the unobserved random draw. Return the conditional probability of each candidate event given the bag. These events are mutually exclusive and exhaustive.", "criteria": {"yellow": "The sampled ball is yellow", "red": "The sampled ball is red", "green": "The sampled ball is green", "pink": "The sampled ball is pink", "cyan": "The sampled ball is cyan", "purple": "The sampled ball is purple", "white": "The sampled ball is white", "orange": "The sampled ball is orange"}}, "focus": {"type": "boolean", "instructions": "Will the unobserved draw produce a white ball? Use the conditional probability given only this bag."}, "group": {"type": "score", "instructions": "Predict which mutually exclusive group the unobserved sampled color belongs to. Probabilities are determined by the bag counts.", "criteria": ["The drawn color is one of: cyan, pink, red", "The drawn color is one of: green, yellow, purple", "The drawn color is one of: orange, white"]}}, "gold": {}, "gold_probs": {"outcome": {"yellow": 0.16666666666666666, "red": 0.14583333333333334, "green": 0.10416666666666667, "pink": 0.14583333333333334, "cyan": 0.020833333333333332, "purple": 0.125, "white": 0.10416666666666667, "orange": 0.1875}, "focus": {"false": 0.8958333333333334, "true": 0.10416666666666667}, "group": {"0": 0.3125, "1": 0.3958333333333333, "2": 0.2916666666666667}}, "gold_probs_kind": "programmatic_conditional_distribution", "gold_label_kind": "unobserved", "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "chance:[[\"cyan\",\"1/48\"],[\"green\",\"5/48\"],[\"orange\",\"3/16\"],[\"pink\",\"7/48\"],[\"purple\",\"1/8\"],[\"red\",\"7/48\"],[\"white\",\"5/48\"],[\"yellow\",\"1/6\"]]", "template_id": "known_chance_v2:ood"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"outcome": {"green": 0, "white": 0, "yellow": 0.01, "purple": 0.01, "red": 0.01, "orange": 0.86, "pink": 0.01, "cyan": 0.1}, "focus": {"false": 0.87, "true": 0.13}, "group": {"0": 0.09, "1": 0.16, "2": 0.75}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:3fc225c57c9937052a49d897", "state_id": "grid_navigation:3fc225c57c9937052a49d897", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . .\n. . . .\n. # . A\n# # . G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "west": 0.0, "south": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (0, 3), (1, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [2, 1], [3, 0], [3, 1]], "position": [2, 3], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.31, "north": 0.12, "south": 0.5700000000000001}, "solvable": {"false": 0.16000000000000003, "true": 0.84}, "value": {"0": 0.64, "1": 0.25, "2": 0.07, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:cb58dff16610adb8da0d56a1", "state_id": "grid_navigation:cb58dff16610adb8da0d56a1", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\nA . # .\n. # . G\n# . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.3333333333333333, "east": 0.3333333333333333, "south": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (1, 2), (2, 1), (2, 3), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 2], [2, 1], [3, 0], [3, 2]], "position": [1, 0], "goal": [2, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.1, "north": 0.09, "east": 0.81}, "solvable": {"false": 0.43000000000000005, "true": 0.57}, "value": {"0": 0.01, "1": 0.4, "2": 0.37, "3": 0.22}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:71198246332dec8c139479eb", "state_id": "tic_tac_toe:71198246332dec8c139479eb", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O X\nX . X\n. O O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.5, "cell_5": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOX.XXOX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOXX.X.OO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.61, "cell_7": 0.39}, "solvable": {"false": 0.85, "true": 0.15}, "value": {"0": 0.58, "1": 0.26, "2": 0.16}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:8c7037241fa142a83064b2bf", "state_id": "tic_tac_toe:8c7037241fa142a83064b2bf", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X .\nX . O\nO X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.0, "cell_3": 0.0, "cell_5": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.O.X.XOXO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OX.X.OOX.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.29, "cell_3": 0.23, "cell_5": 0.48}, "solvable": {"false": 0.5900000000000001, "true": 0.41}, "value": {"0": 0.17, "1": 0.34, "2": 0.49}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:351bd47c4b79b72ad97cf473", "state_id": "tic_tac_toe:351bd47c4b79b72ad97cf473", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X .\n. . .\nX O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_3": "Place X in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_4": 0.2, "cell_9": 0.2, "cell_6": 0.2, "cell_5": 0.2, "cell_3": 0.2}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_4", "cell_5", "cell_6", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...O.XX.O:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OX....XO.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.17, "cell_5": 0.39, "cell_6": 0.11, "cell_9": 0.21, "cell_4": 0.12}, "solvable": {"false": 0.7, "true": 0.3}, "value": {"0": 0.15, "1": 0.51, "2": 0.34}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:bb113d6d28db0e760e635b08", "state_id": "grid_navigation:bb113d6d28db0e760e635b08", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG . . .\n. . . .\n. # # .\nA . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((1, 2), (2, 2), (2, 3), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[2, 1], [2, 2], [3, 2], [3, 3]], "position": [3, 0], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.92, "east": 0.08}, "solvable": {"false": 0.22999999999999998, "true": 0.77}, "value": {"0": 0.01, "1": 0.39, "2": 0.5, "3": 0.1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:d0718319f602dbbed77f888d", "state_id": "grid_navigation:d0718319f602dbbed77f888d", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . A\n. # # .\nG . . #\n# . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": false, "value": 3}, "gold_probs": {"action": {"south": 0.5, "west": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (1, 2), (1, 3), (2, 2), (3, 0), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 1], [1, 2], [2, 3], [3, 0], [3, 3]], "position": [0, 3], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.39, "south": 0.61}, "solvable": {"false": 0.31000000000000005, "true": 0.69}, "value": {"0": 0.01, "1": 0.44, "2": 0.43, "3": 0.12}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:0aa66fc9f37e5bba368459f0", "state_id": "tic_tac_toe:0aa66fc9f37e5bba368459f0", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O X\n. O .\n. X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_4": "Place X in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_7": 0.3333333333333333, "cell_6": 0.3333333333333333, "cell_4": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4", "cell_6", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XXOOO.X:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOX.O..XO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.18, "cell_4": 0.53, "cell_7": 0.29}, "solvable": {"false": 0.8, "true": 0.2}, "value": {"0": 0.35, "1": 0.43, "2": 0.22}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:14e57129e39f50b1868380db", "state_id": "tic_tac_toe:14e57129e39f50b1868380db", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\nO O X\nX . X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place O in row 3, column 2 (cell 8).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_8": 0.25, "cell_2": 0.25, "cell_1": 0.25, "cell_3": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_2", "cell_3", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...OOXX.X:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "...OOXX.X", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.2, "cell_8": 0.56, "cell_3": 0.12, "cell_1": 0.12}, "solvable": {"false": 0.83, "true": 0.17}, "value": {"0": 0.54, "1": 0.32, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:37d194c31d160acba7263c83", "state_id": "tic_tac_toe:37d194c31d160acba7263c83", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O O\n. X X\n. X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_4": "Place O in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_7": 0.0, "cell_4": 1.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XXXOOXO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOO.XX.XO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.42, "cell_4": 0.58}, "solvable": {"false": 0.86, "true": 0.14}, "value": {"0": 0.47, "1": 0.39, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:01f683d11a99cd2f1974ad8d", "state_id": "grid_navigation:01f683d11a99cd2f1974ad8d", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A . #\n. G . .\n. . . #\n# # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 1.0, "west": 0.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 3), (1, 3), (2, 3), (3, 0), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [2, 3], [3, 0], [3, 1], [3, 2]], "position": [0, 1], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.01, "south": 0.68, "west": 0.31}, "solvable": {"false": 0.06999999999999995, "true": 0.93}, "value": {"0": 0.94, "1": 0.04, "2": 0.01, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:be00d2ffe1fa3888aa672faa", "state_id": "grid_navigation:be00d2ffe1fa3888aa672faa", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A . .\n# # . .\n. # . .\nG . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "west": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (1, 1), (1, 2), (2, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [1, 1], [2, 1], [3, 2]], "position": [0, 1], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.67, "east": 0.33}, "solvable": {"false": 0.36, "true": 0.64}, "value": {"0": 0.01, "1": 0.38, "2": 0.48, "3": 0.13}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:0268bc49206669de0c03a52a", "state_id": "tic_tac_toe:0268bc49206669de0c03a52a", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . X\nO X .\n. O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_6": "Place O in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_7": 0.25, "cell_1": 0.25, "cell_2": 0.25, "cell_6": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_2", "cell_6", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOX..OX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..XOX..OX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.27, "cell_2": 0.24, "cell_6": 0.33, "cell_7": 0.16}, "solvable": {"false": 0.92, "true": 0.08}, "value": {"0": 0.67, "1": 0.23, "2": 0.1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:9c18c218d5f9baeeca41b722", "state_id": "grid_navigation:9c18c218d5f9baeeca41b722", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A . .\n. . . #\n# . . .\n. . . G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.5, "east": 0.5, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 3], [2, 0]], "position": [0, 1], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.44, "east": 0.54, "west": 0.02}, "solvable": {"false": 0.07999999999999996, "true": 0.92}, "value": {"0": 0, "1": 0.4, "2": 0.56, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:90394f92f11a46aecee283bb", "state_id": "tic_tac_toe:90394f92f11a46aecee283bb", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . X\nX X O\n. O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_7", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_7": 1.0, "cell_2": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXXXOO.X:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.XXXO.OX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.21, "cell_7": 0.79}, "solvable": {"false": 0.88, "true": 0.12}, "value": {"0": 0.46, "1": 0.4, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:8f75ed644ca4e5ad38876d38", "state_id": "tic_tac_toe:8f75ed644ca4e5ad38876d38", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . X\nX . O\nX . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place O in row 1, column 2 (cell 2).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_2": 0.0, "cell_8": 0.0, "cell_9": 0.0, "cell_5": 1.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XO.XX.O:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.XX.OX..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.26, "cell_9": 0.16, "cell_2": 0.2, "cell_5": 0.38}, "solvable": {"false": 0.86, "true": 0.14}, "value": {"0": 0.62, "1": 0.21, "2": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:a4da1c8fb7700fc624ccfe50", "state_id": "grid_navigation:a4da1c8fb7700fc624ccfe50", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # .\n# # # G\n. # . A\n. . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "north": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (1, 1), (1, 3), (2, 1), (2, 2), (3, 0), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [1, 0], [1, 1], [1, 2], [2, 1], [3, 2]], "position": [2, 3], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.26, "south": 0.14, "north": 0.6}, "solvable": {"false": 0.53, "true": 0.47}, "value": {"0": 0.15, "1": 0.37, "2": 0.19, "3": 0.29}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:cafb22acaf6ff37b738c75a8", "state_id": "tic_tac_toe:cafb22acaf6ff37b738c75a8", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X .\nO X X\n. O .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_7", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 1.0, "cell_9": 0.0, "cell_1": 0.0, "cell_3": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.O.OXX.X.:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".X.OXX.O.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.23, "cell_3": 0.32, "cell_9": 0.17, "cell_1": 0.28}, "solvable": {"false": 0.83, "true": 0.17}, "value": {"0": 0.46, "1": 0.3, "2": 0.24}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:ff0a80f3cd183b2ddc7739a5", "state_id": "grid_navigation:ff0a80f3cd183b2ddc7739a5", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . G\n. # # #\n# . # .\n# A . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (1, 1), (1, 2), (2, 1), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 1], [1, 2], [1, 3], [2, 0], [2, 2], [3, 0]], "position": [3, 1], "goal": [0, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.61, "east": 0.39}, "solvable": {"false": 0.6, "true": 0.4}, "value": {"0": 0.02, "1": 0.31, "2": 0.35, "3": 0.32}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:4314702d76009bb85c8b1d5c", "state_id": "tic_tac_toe:4314702d76009bb85c8b1d5c", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X O\nX X O\n. . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_8", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.0, "cell_8": 1.0, "cell_7": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...OXXOXO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXOXXO...", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.15, "cell_8": 0.14, "cell_7": 0.71}, "solvable": {"false": 0.64, "true": 0.36}, "value": {"0": 0.12, "1": 0.37, "2": 0.51}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:0208ce6a665e9e36394b662e", "state_id": "tic_tac_toe:0208ce6a665e9e36394b662e", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . O\n. X X\n. O X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_4": "Place X in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.0, "cell_2": 0.0, "cell_4": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OOX.XXO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.O.XX.OX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.21, "cell_4": 0.35, "cell_7": 0.44}, "solvable": {"false": 0.51, "true": 0.49}, "value": {"0": 0.21, "1": 0.27, "2": 0.52}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:9c48cbf99b57a593a04bc08a", "state_id": "tic_tac_toe:9c48cbf99b57a593a04bc08a", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO O .\nX . .\nO X X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place X in row 1, column 3 (cell 3).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_3": 1.0, "cell_6": 0.0, "cell_5": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XO.XOXO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OO.X..OXX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.22, "cell_3": 0.34, "cell_5": 0.44}, "solvable": {"false": 0.55, "true": 0.45}, "value": {"0": 0.24, "1": 0.27, "2": 0.49}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:ca31d6c52e7a647a94ba8ee9", "state_id": "tic_tac_toe:ca31d6c52e7a647a94ba8ee9", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO O .\nO X .\nX X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place X in row 1, column 3 (cell 3).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_3": 0.5, "cell_9": 0.5, "cell_6": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...OXXOOX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OO.OX.XX.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.32, "cell_3": 0.28, "cell_9": 0.39999999999999997}, "solvable": {"false": 0.44999999999999996, "true": 0.55}, "value": {"0": 0.24, "1": 0.23, "2": 0.53}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:a247e19b534dea6d3f496f89", "state_id": "tic_tac_toe:a247e19b534dea6d3f496f89", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O .\nX X .\n. O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_1": "Place X in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.2, "cell_7": 0.2, "cell_3": 0.2, "cell_6": 0.2, "cell_1": 0.2}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_3", "cell_6", "cell_7", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...OXO.X.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".O.XX..O.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.18, "cell_1": 0.13, "cell_7": 0.11, "cell_6": 0.51, "cell_9": 0.07}, "solvable": {"false": 0.5800000000000001, "true": 0.42}, "value": {"0": 0.21, "1": 0.38, "2": 0.41}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:96c816ec81d12843705b6df8", "state_id": "grid_navigation:96c816ec81d12843705b6df8", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . A\nG # . .\n# # . .\n# . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (0, 3), (1, 1), (1, 2), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 1], [2, 0], [2, 1], [3, 0], [3, 2]], "position": [0, 3], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.4, "west": 0.6}, "solvable": {"false": 0.41000000000000003, "true": 0.59}, "value": {"0": 0.04, "1": 0.46, "2": 0.31, "3": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:fe910a10992dfcb3929d7f6b", "state_id": "grid_navigation:fe910a10992dfcb3929d7f6b", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . G\n# . . #\n. . # .\n# # . A", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.5, "west": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (1, 2), (2, 3), (3, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [1, 3], [2, 2], [3, 0], [3, 1]], "position": [3, 3], "goal": [0, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.84, "west": 0.16}, "solvable": {"false": 0.31000000000000005, "true": 0.69}, "value": {"0": 0.01, "1": 0.42, "2": 0.43, "3": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:38bce347f09ebd6f8c77b6ba", "state_id": "grid_navigation:38bce347f09ebd6f8c77b6ba", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\n. # # .\n. # # #\nA . G #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (1, 0), (1, 1), (1, 2), (2, 1), (2, 2), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 1], [1, 2], [2, 1], [2, 2], [2, 3], [3, 3]], "position": [3, 0], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.29, "east": 0.71}, "solvable": {"false": 0.32999999999999996, "true": 0.67}, "value": {"0": 0.62, "1": 0.18, "2": 0.12, "3": 0.08}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:468de14ecaa0b47f5b0ea1bb", "state_id": "tic_tac_toe:468de14ecaa0b47f5b0ea1bb", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . X\nX . .\n. O O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place X in row 2, column 2 (cell 5).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_7", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 0.0, "cell_7": 1.0, "cell_1": 0.0, "cell_6": 0.0, "cell_2": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XX...OO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..XX...OO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.15, "cell_7": 0.27, "cell_5": 0.36, "cell_1": 0.1, "cell_2": 0.12}, "solvable": {"false": 0.5700000000000001, "true": 0.43}, "value": {"0": 0.26, "1": 0.35, "2": 0.39}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:52796f350c689d5b33c91255", "state_id": "grid_navigation:52796f350c689d5b33c91255", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # .\n# # # G\n# A . #\n# . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "south": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (0, 3), (1, 0), (1, 1), (1, 3), (2, 1), (3, 0), (3, 1), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [1, 0], [1, 1], [1, 2], [2, 0], [2, 3], [3, 0], [3, 2], [3, 3]], "position": [2, 1], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.85, "south": 0.15}, "solvable": {"false": 0.54, "true": 0.46}, "value": {"0": 0.2, "1": 0.33, "2": 0.08, "3": 0.39}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:a468f0be302f0f8e35065cba", "state_id": "grid_navigation:a468f0be302f0f8e35065cba", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nA . . .\nG . # .\n. . . .\n# . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.0, "south": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (2, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 2], [3, 0]], "position": [0, 0], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.82, "east": 0.18}, "solvable": {"false": 0.10999999999999999, "true": 0.89}, "value": {"0": 0.5, "1": 0.43, "2": 0.06, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:7b2014c32cdb710e684020dd", "state_id": "tic_tac_toe:7b2014c32cdb710e684020dd", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X X\n. . .\nO O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place O in row 2, column 3 (cell 6).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_4": "Place O in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_6": 0.25, "cell_5": 0.25, "cell_1": 0.25, "cell_4": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_4", "cell_5", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OX.OX.X:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XX...OOX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.28, "cell_5": 0.28, "cell_6": 0.16, "cell_1": 0.28}, "solvable": {"false": 0.81, "true": 0.19}, "value": {"0": 0.35, "1": 0.36, "2": 0.29}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:914064587ff4007faeaeaba6", "state_id": "tic_tac_toe:914064587ff4007faeaeaba6", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . X\nO X O\n. O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place X in row 1, column 2 (cell 2).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.3333333333333333, "cell_9": 0.3333333333333333, "cell_7": 0.3333333333333333}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_7", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.O.OXOX.X:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.XOXO.O.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.25, "cell_7": 0.52, "cell_2": 0.23}, "solvable": {"false": 0.6, "true": 0.4}, "value": {"0": 0.14, "1": 0.38, "2": 0.48}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:382a81484c4e1ddc6130d3bb", "state_id": "grid_navigation:382a81484c4e1ddc6130d3bb", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\nA G . .\n. . # .\n# . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "east": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (2, 2), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [2, 2], [3, 0]], "position": [1, 0], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.06, "north": 0.21, "east": 0.73}, "solvable": {"false": 0.040000000000000036, "true": 0.96}, "value": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:be64bd991ed9a443b87b6064", "state_id": "tic_tac_toe:be64bd991ed9a443b87b6064", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X O\n. . .\nO . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place X in row 2, column 3 (cell 6).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.0, "cell_5": 1.0, "cell_4": 0.0, "cell_8": 0.0, "cell_9": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O...OXX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XXO...O..", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.09, "cell_6": 0.08, "cell_4": 0.19, "cell_9": 0.1, "cell_5": 0.54}, "solvable": {"false": 0.75, "true": 0.25}, "value": {"0": 0.18, "1": 0.5, "2": 0.32}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:81c0dd7075bf44b499231769", "state_id": "tic_tac_toe:81c0dd7075bf44b499231769", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO O .\n. . X\nX X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place O in row 1, column 3 (cell 3).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_3": 0.5, "cell_5": 0.0, "cell_4": 0.0, "cell_9": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOX...XX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OO...XXX.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.12, "cell_3": 0.17, "cell_5": 0.22, "cell_9": 0.49}, "solvable": {"false": 0.78, "true": 0.22}, "value": {"0": 0.43, "1": 0.33, "2": 0.24}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:e2492900d6d8ee9bda93ac8e", "state_id": "grid_navigation:e2492900d6d8ee9bda93ac8e", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # # .\n. A # G\n. . . .\n. # . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 0.0, "south": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 0), (1, 1), (2, 0), (2, 3), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [0, 2], [1, 2], [3, 1], [3, 3]], "position": [1, 1], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.85, "west": 0.15}, "solvable": {"false": 0.5900000000000001, "true": 0.41}, "value": {"0": 0.23, "1": 0.48, "2": 0.12, "3": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:06ca74c571acdc6a345c9973", "state_id": "grid_navigation:06ca74c571acdc6a345c9973", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . #\n. # # #\n. G . .\n# A . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 1.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (1, 2), (2, 2), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 1], [1, 2], [1, 3], [3, 0]], "position": [3, 1], "goal": [2, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.36, "north": 0.64}, "solvable": {"false": 0.30000000000000004, "true": 0.7}, "value": {"0": 0.33, "1": 0.43, "2": 0.13, "3": 0.11}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:eff62ca1a568cda284569094", "state_id": "grid_navigation:eff62ca1a568cda284569094", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . A .\n# . # .\n. . # .\n. . G .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 0.5, "east": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 1), (2, 1), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [1, 2], [2, 2]], "position": [0, 2], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.64, "east": 0.36}, "solvable": {"false": 0.4, "true": 0.6}, "value": {"0": 0.04, "1": 0.54, "2": 0.3, "3": 0.12}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:c409c1117582d89a74a4769f", "state_id": "grid_navigation:c409c1117582d89a74a4769f", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. . A G\n. . . #\n. # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south.", "east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "south": 0.0, "east": 1.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 3), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[2, 3], [3, 1], [3, 2]], "position": [1, 2], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0, "west": 0.01, "north": 0, "east": 0.99}, "solvable": {"false": 0.030000000000000027, "true": 0.97}, "value": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:e5fea558c983409e424c0f19", "state_id": "grid_navigation:e5fea558c983409e424c0f19", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . .\n# # G #\n. . # #\n# A . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (0, 2), (0, 3), (1, 2), (2, 1), (3, 0), (3, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 0], [1, 1], [1, 3], [2, 2], [2, 3], [3, 0], [3, 3]], "position": [3, 1], "goal": [1, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.7, "east": 0.3}, "solvable": {"false": 0.54, "true": 0.46}, "value": {"0": 0.08, "1": 0.44, "2": 0.17, "3": 0.31}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:22524ab1bf5bcdf516a3499f", "state_id": "grid_navigation:22524ab1bf5bcdf516a3499f", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# G . .\n. . . A\n# . . .\n. # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 0.5, "south": 0.0, "north": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (2, 0), (3, 1), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [2, 0], [3, 1], [3, 2]], "position": [1, 3], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.44, "south": 0.02, "west": 0.54}, "solvable": {"false": 0.18999999999999995, "true": 0.81}, "value": {"0": 0.06, "1": 0.71, "2": 0.18, "3": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:856ff071ec6f9824675c94a9", "state_id": "grid_navigation:856ff071ec6f9824675c94a9", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nA . # #\n. . # #\n# . # .\n# G . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.5, "east": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (1, 0), (1, 2), (2, 2), (2, 3), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [1, 2], [1, 3], [2, 0], [2, 2], [3, 0]], "position": [0, 0], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.03, "south": 0.97}, "solvable": {"false": 0.19999999999999996, "true": 0.8}, "value": {"0": 0.03, "1": 0.63, "2": 0.25, "3": 0.09}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:8af78c5be848e6bef7ea495a", "state_id": "grid_navigation:8af78c5be848e6bef7ea495a", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # .\n. # # #\n. . G A\n# # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (0, 2), (0, 3), (2, 0), (2, 1), (2, 2), (3, 1), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [1, 1], [1, 2], [1, 3], [3, 0], [3, 1], [3, 2]], "position": [2, 3], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.97, "south": 0.03}, "solvable": {"false": 0.08999999999999997, "true": 0.91}, "value": {"0": 0.89, "1": 0.07, "2": 0.02, "3": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:eeab2db63394e68e9014f390", "state_id": "tic_tac_toe:eeab2db63394e68e9014f390", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O .\nX . O\n. X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place X in row 1, column 3 (cell 3).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_7", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_3": 0.0, "cell_7": 1.0, "cell_5": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOO.XXX.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XO.X.O.XO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.55, "cell_3": 0.19, "cell_7": 0.26}, "solvable": {"false": 0.76, "true": 0.24}, "value": {"0": 0.41, "1": 0.37, "2": 0.22}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:ea9f2a734a2e6e2dc0593838", "state_id": "tic_tac_toe:ea9f2a734a2e6e2dc0593838", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X X\nX . .\nO X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place O in row 2, column 2 (cell 5).", "cell_6": "Place O in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 1.0, "cell_6": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.XX.XOXO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXXX..OXO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.26, "cell_5": 0.74}, "solvable": {"false": 0.84, "true": 0.16}, "value": {"0": 0.49, "1": 0.32, "2": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:64d79cc0343bae4850bb7a48", "state_id": "tic_tac_toe:64d79cc0343bae4850bb7a48", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O .\nX . X\nO O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place X in row 2, column 2 (cell 5).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 1.0, "cell_3": 0.0, "cell_9": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOX.X.OX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XO.X.XOO.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.32, "cell_9": 0.3, "cell_5": 0.38}, "solvable": {"false": 0.47, "true": 0.53}, "value": {"0": 0.15, "1": 0.22, "2": 0.63}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:b5da5a72759718738d8cd070", "state_id": "grid_navigation:b5da5a72759718738d8cd070", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . #\n# . # .\n. # . A\nG . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.5, "west": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (0, 3), (1, 1), (2, 0), (2, 2), (3, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 3], [1, 0], [1, 2], [2, 1], [3, 2], [3, 3]], "position": [2, 3], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.32, "west": 0.68}, "solvable": {"false": 0.43999999999999995, "true": 0.56}, "value": {"0": 0.05, "1": 0.44, "2": 0.25, "3": 0.26}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:d5e33aa7ad91b281411b4c48", "state_id": "tic_tac_toe:d5e33aa7ad91b281411b4c48", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . X\nO . O\nX X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place X in row 2, column 2 (cell 5).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 1.0, "cell_1": 0.0, "cell_2": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XO.OXXO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..XO.OXXO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.22, "cell_2": 0.24, "cell_5": 0.54}, "solvable": {"false": 0.62, "true": 0.38}, "value": {"0": 0.15, "1": 0.38, "2": 0.47}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:aa4f8444fdd51babd60fafc7", "state_id": "grid_navigation:aa4f8444fdd51babd60fafc7", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nA . . .\n. . . .\n. . # #\n. . G .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.5, "east": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (1, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[2, 2], [2, 3]], "position": [0, 0], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.83, "east": 0.17}, "solvable": {"false": 0.09999999999999998, "true": 0.9}, "value": {"0": 0.01, "1": 0.53, "2": 0.42, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:a1d0df6d9d205bf126467cbc", "state_id": "tic_tac_toe:a1d0df6d9d205bf126467cbc", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X .\nO . O\nX . O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place X in row 3, column 2 (cell 8).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_8": 0.0, "cell_3": 1.0, "cell_5": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOX..XOX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XX.O.OX.O", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.41, "cell_8": 0.21, "cell_5": 0.38}, "solvable": {"false": 0.48, "true": 0.52}, "value": {"0": 0.23, "1": 0.3, "2": 0.47}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:e5c3f95adba0d60db6d2407d", "state_id": "tic_tac_toe:e5c3f95adba0d60db6d2407d", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X X\nO X X\n. O .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_7": "Place O in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_7", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.0, "cell_7": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.O.OXXOXX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXXOXX.O.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.78, "cell_9": 0.22}, "solvable": {"false": 0.84, "true": 0.16}, "value": {"0": 0.48, "1": 0.31, "2": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:622b8dbbfa014ebe7c23a8cf", "state_id": "tic_tac_toe:622b8dbbfa014ebe7c23a8cf", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . O\n. . X\nX O O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place X in row 1, column 2 (cell 2).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_4": "Place X in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.0, "cell_5": 0.0, "cell_4": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.XX..OOX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.O..XXOO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.33, "cell_2": 0.22, "cell_5": 0.45}, "solvable": {"false": 0.74, "true": 0.26}, "value": {"0": 0.22, "1": 0.47, "2": 0.31}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:57fc8eaade31d2d7ebfd50a0", "state_id": "grid_navigation:57fc8eaade31d2d7ebfd50a0", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. A # .\n. . . G\n. # . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 0.0, "north": 0.0, "south": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (1, 2), (2, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 2], [3, 1], [3, 3]], "position": [1, 1], "goal": [2, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.15, "west": 0.39, "south": 0.46}, "solvable": {"false": 0.18000000000000005, "true": 0.82}, "value": {"0": 0.21, "1": 0.66, "2": 0.08, "3": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:a8e9aff98c344c128cef045b", "state_id": "grid_navigation:a8e9aff98c344c128cef045b", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\n. . . .\n. . . .\nA G . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((3, 1),)", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1]], "position": [3, 0], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.52, "east": 0.48}, "solvable": {"false": 0.030000000000000027, "true": 0.97}, "value": {"0": 0.99, "1": 0.01, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:61bba4699cd7fb123e97e367", "state_id": "grid_navigation:61bba4699cd7fb123e97e367", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # #\n. . . #\n. . . A\n. . G .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.5, "south": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((2, 0), (3, 0), (3, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [0, 3], [1, 3]], "position": [2, 3], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.65, "west": 0.35}, "solvable": {"false": 0.07999999999999996, "true": 0.92}, "value": {"0": 0.6, "1": 0.36, "2": 0.03, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:26753943e93a064ed40bd82d", "state_id": "grid_navigation:26753943e93a064ed40bd82d", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # .\nA . # .\n. . G #\n. # # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.0, "east": 0.5, "south": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 1), (0, 2), (1, 0), (2, 1), (3, 1), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [1, 2], [2, 3], [3, 1], [3, 2], [3, 3]], "position": [1, 0], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.57, "north": 0.03, "south": 0.4}, "solvable": {"false": 0.13, "true": 0.87}, "value": {"0": 0.29, "1": 0.58, "2": 0.07, "3": 0.06}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:220e9bc0e7240848b57b466b", "state_id": "tic_tac_toe:220e9bc0e7240848b57b466b", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O O\nO X X\n. X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_7": 0.0, "cell_9": 1.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXXXO.XO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOOOXX.X.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.37, "cell_7": 0.63}, "solvable": {"false": 0.81, "true": 0.19}, "value": {"0": 0.33, "1": 0.44, "2": 0.23}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:b1ca09fa0b455f900245f809", "state_id": "grid_navigation:b1ca09fa0b455f900245f809", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . #\n# . # #\n. . A G\n# # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 1.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (0, 3), (1, 0), (1, 2), (2, 0), (2, 3), (3, 0), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 3], [1, 0], [1, 2], [1, 3], [3, 0], [3, 1], [3, 2]], "position": [2, 2], "goal": [2, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.05, "east": 0.95}, "solvable": {"false": 0.06000000000000005, "true": 0.94}, "value": {"0": 0.99, "1": 0, "2": 0, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:fbacaab841f1d108a8d14436", "state_id": "grid_navigation:fbacaab841f1d108a8d14436", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # #\n. A # .\n# # # .\n. # # G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.5, "west": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 2), (1, 1), (1, 2), (1, 3), (2, 1), (3, 0), (3, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [0, 3], [1, 2], [2, 0], [2, 1], [2, 2], [3, 1], [3, 2]], "position": [1, 1], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.73, "north": 0.27}, "solvable": {"false": 0.6, "true": 0.4}, "value": {"0": 0.01, "1": 0.34, "2": 0.34, "3": 0.31}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:45fad76f12f4876ae56ad9de", "state_id": "tic_tac_toe:45fad76f12f4876ae56ad9de", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X X\nX O O\n. O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 1.0, "cell_7": 0.0, "cell_9": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.O.OOXXX.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XXXOO.O.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.34, "cell_9": 0.3, "cell_1": 0.36}, "solvable": {"false": 0.55, "true": 0.45}, "value": {"0": 0.24, "1": 0.31, "2": 0.45}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:4c248c08ce64b1d5dbe24040", "state_id": "tic_tac_toe:4c248c08ce64b1d5dbe24040", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X .\nO O X\n. X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_7": "Place O in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.5, "cell_3": 0.0, "cell_9": 0.0, "cell_7": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.O.XOX.X.:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".X.OOX.X.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.14, "cell_3": 0.18, "cell_1": 0.41, "cell_7": 0.27}, "solvable": {"false": 0.8200000000000001, "true": 0.18}, "value": {"0": 0.43, "1": 0.36, "2": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:07f8603a9dba55c19fcd49f2", "state_id": "tic_tac_toe:07f8603a9dba55c19fcd49f2", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O O\n. . .\nX . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_4": "Place X in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.2, "cell_5": 0.2, "cell_8": 0.2, "cell_6": 0.2, "cell_4": 0.2}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4", "cell_5", "cell_6", "cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O..OX.X:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOO...X..", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.35, "cell_5": 0.41, "cell_9": 0.08, "cell_8": 0.11, "cell_6": 0.05}, "solvable": {"false": 0.79, "true": 0.21}, "value": {"0": 0.23, "1": 0.44, "2": 0.33}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:d18ab4f3b93ee36c6b8bfc91", "state_id": "tic_tac_toe:d18ab4f3b93ee36c6b8bfc91", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X .\nX . O\nO X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place O in row 1, column 3 (cell 3).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_3": 1.0, "cell_5": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOX.XXXO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XX.X.OOXO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.38, "cell_5": 0.62}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.57, "1": 0.28, "2": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:9d2a8896cd9de0537704bf89", "state_id": "grid_navigation:9d2a8896cd9de0537704bf89", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . #\n# A . #\n# . G #\n. . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "south": 0.5, "east": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (0, 2), (0, 3), (1, 0), (3, 1), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 0], [1, 3], [2, 0], [2, 3], [3, 2]], "position": [1, 1], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.02, "south": 0.7, "east": 0.28}, "solvable": {"false": 0.15000000000000002, "true": 0.85}, "value": {"0": 0.87, "1": 0.07, "2": 0.02, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:d172b7697707b5307245d334", "state_id": "tic_tac_toe:d172b7697707b5307245d334", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X X\nO X .\n. O .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_1": 0.25, "cell_6": 0.25, "cell_7": 0.25, "cell_9": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_6", "cell_7", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOXX.O.:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XXOX..O.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.26, "cell_6": 0.33, "cell_7": 0.18, "cell_1": 0.23}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.51, "1": 0.33, "2": 0.16}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:cc283c619145b6b0e8471ba8", "state_id": "tic_tac_toe:cc283c619145b6b0e8471ba8", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X O\nX . X\n. . O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_8": "Place O in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.0, "cell_1": 0.0, "cell_5": 1.0, "cell_8": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OX.X.XO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XOX.X..O", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.2, "cell_7": 0.18, "cell_5": 0.5, "cell_8": 0.12}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.45, "1": 0.38, "2": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:bfced1690164da8e81ceb841", "state_id": "grid_navigation:bfced1690164da8e81ceb841", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG . # .\n# . # A\n# . # .\n. . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.5, "south": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 2), (2, 0), (2, 1), (2, 2), (2, 3), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 0], [1, 2], [2, 0], [2, 2], [3, 2], [3, 3]], "position": [1, 3], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.62, "south": 0.38}, "solvable": {"false": 0.47, "true": 0.53}, "value": {"0": 0.06, "1": 0.36, "2": 0.28, "3": 0.3}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:1fbb47398a6b9049a907b971", "state_id": "tic_tac_toe:1fbb47398a6b9049a907b971", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . O\n. O X\nX . X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place O in row 3, column 2 (cell 8).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_1": "Place O in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_8", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_8": 1.0, "cell_4": 0.0, "cell_2": 0.0, "cell_1": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O.OXX.X:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..O.OXX.X", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.14, "cell_4": 0.25, "cell_2": 0.19, "cell_8": 0.42}, "solvable": {"false": 0.8200000000000001, "true": 0.18}, "value": {"0": 0.32, "1": 0.41, "2": 0.27}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:0bcd75f9277e862aae188399", "state_id": "grid_navigation:0bcd75f9277e862aae188399", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . A .\nG # . #\n. # . #\n# # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.0, "east": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 1), (1, 2), (1, 3), (2, 3), (3, 1), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 1], [1, 3], [2, 1], [2, 3], [3, 0], [3, 1], [3, 2]], "position": [0, 2], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.14, "west": 0.77, "south": 0.09}, "solvable": {"false": 0.5, "true": 0.5}, "value": {"0": 0.09, "1": 0.42, "2": 0.22, "3": 0.27}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:8c90af2bd687c4037a0698f2", "state_id": "grid_navigation:8c90af2bd687c4037a0698f2", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # .\n# # . #\n# A . .\nG . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.0, "south": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 2), (0, 3), (1, 2), (2, 0), (2, 3), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [1, 0], [1, 1], [1, 3], [2, 0], [3, 2]], "position": [2, 1], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.83, "east": 0.17}, "solvable": {"false": 0.27, "true": 0.73}, "value": {"0": 0.36, "1": 0.45, "2": 0.12, "3": 0.07}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:2752f5ded809fdbdc91c7002", "state_id": "tic_tac_toe:2752f5ded809fdbdc91c7002", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X X\nX . O\nX O O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place O in row 2, column 2 (cell 5).", "cell_1": "Place O in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_5": 0.5, "cell_1": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.XXX.OXOO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XXX.OXOO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.62, "cell_1": 0.38}, "solvable": {"false": 0.83, "true": 0.17}, "value": {"0": 0.6, "1": 0.23, "2": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:4927ff833f607c87c1d8221b", "state_id": "tic_tac_toe:4927ff833f607c87c1d8221b", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X O\nX . O\nO X X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.0, "cell_5": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.XOX.OOXX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XOX.OOXX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.35, "cell_5": 0.65}, "solvable": {"false": 0.85, "true": 0.15}, "value": {"0": 0.43, "1": 0.41, "2": 0.16}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:e7842a63cfee0f5b711b20be", "state_id": "grid_navigation:e7842a63cfee0f5b711b20be", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . #\n. . . .\n. G . A\n. # . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (0, 3), (3, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 3], [3, 1], [3, 3]], "position": [2, 3], "goal": [2, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.85, "north": 0.15}, "solvable": {"false": 0.07999999999999996, "true": 0.92}, "value": {"0": 0.4, "1": 0.52, "2": 0.07, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:5ecc09c9a881d1d89d8a6c76", "state_id": "grid_navigation:5ecc09c9a881d1d89d8a6c76", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . #\n. . # .\n. . G A\n# . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "west": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 3), (1, 2), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 2], [3, 0]], "position": [2, 3], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.05, "south": 0, "west": 0.95}, "solvable": {"false": 0.09999999999999998, "true": 0.9}, "value": {"0": 0.88, "1": 0.07, "2": 0.02, "3": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:ecd3cf909f4686acc5c72344", "state_id": "grid_navigation:ecd3cf909f4686acc5c72344", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nA . . #\n. # . .\n# # . .\n. . G .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 2), (1, 3), (2, 2), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 1], [2, 0], [2, 1]], "position": [0, 0], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.67, "east": 0.33}, "solvable": {"false": 0.31000000000000005, "true": 0.69}, "value": {"0": 0.01, "1": 0.32, "2": 0.57, "3": 0.1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:f5cb2aadaa51e0fff18723e7", "state_id": "tic_tac_toe:f5cb2aadaa51e0fff18723e7", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X O\nX . X\nO . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place O in row 2, column 2 (cell 5).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_8": "Place O in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 1.0, "cell_1": 0.0, "cell_9": 0.0, "cell_8": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OX.XOX.:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XOX.XO..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.2, "cell_5": 0.55, "cell_1": 0.12, "cell_9": 0.13}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.5, "1": 0.31, "2": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:1516b12a9602264b1ea65fa9", "state_id": "tic_tac_toe:1516b12a9602264b1ea65fa9", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X .\n. . X\n. . O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place X in row 1, column 3 (cell 3).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_3": 0.0, "cell_5": 1.0, "cell_4": 0.0, "cell_8": 0.0, "cell_7": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O..XOX.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OX...X..O", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.07, "cell_5": 0.39, "cell_3": 0.38, "cell_4": 0.06, "cell_7": 0.1}, "solvable": {"false": 0.56, "true": 0.44}, "value": {"0": 0.12, "1": 0.35, "2": 0.53}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:0abfa5b9f2bce49f66f35839", "state_id": "grid_navigation:0abfa5b9f2bce49f66f35839", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . G .\n. . # #\n# # # #\n. A . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.5, "east": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 0), (1, 1), (2, 0), (2, 1), (2, 2), (2, 3), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 2], [1, 3], [2, 0], [2, 1], [2, 2], [2, 3], [3, 3]], "position": [3, 1], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.12, "west": 0.88}, "solvable": {"false": 0.39, "true": 0.61}, "value": {"0": 0.03, "1": 0.41, "2": 0.31, "3": 0.25}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:1060dcb59e8fb4b96fbf128d", "state_id": "tic_tac_toe:1060dcb59e8fb4b96fbf128d", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . .\n. O .\nX . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place X in row 1, column 3 (cell 3).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_3": 0.3333333333333333, "cell_8": 0.3333333333333333, "cell_6": 0.3333333333333333, "cell_4": 0.0, "cell_2": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_6", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O.O.X.X:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O...O.X.X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.19, "cell_8": 0.28, "cell_2": 0.15, "cell_6": 0.18, "cell_4": 0.2}, "solvable": {"false": 0.61, "true": 0.39}, "value": {"0": 0.21, "1": 0.36, "2": 0.43}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:09c39bc35251089b8bb5e0f3", "state_id": "grid_navigation:09c39bc35251089b8bb5e0f3", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. . . #\n# . # G\n. # . A", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 1), (2, 0), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 3], [2, 0], [2, 2], [3, 1]], "position": [3, 3], "goal": [2, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.3, "north": 0.7}, "solvable": {"false": 0.38, "true": 0.62}, "value": {"0": 0.11, "1": 0.56, "2": 0.21, "3": 0.12}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:75567fe2239f1424e32593eb", "state_id": "tic_tac_toe:75567fe2239f1424e32593eb", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O X\nO . .\nO X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place X in row 2, column 3 (cell 6).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.0, "cell_5": 0.0, "cell_9": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XX.OOOX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOXO..OX.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.43, "cell_6": 0.23, "cell_9": 0.34}, "solvable": {"false": 0.56, "true": 0.44}, "value": {"0": 0.24, "1": 0.32, "2": 0.44}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:dd10e86576fef2a6a052cd89", "state_id": "grid_navigation:dd10e86576fef2a6a052cd89", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\n. # . G\n# . A #\n. # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 1.0, "west": 0.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (2, 0), (2, 1), (2, 3), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 1], [2, 0], [2, 3], [3, 1]], "position": [2, 2], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.73, "west": 0.16, "south": 0.11}, "solvable": {"false": 0.38, "true": 0.62}, "value": {"0": 0.18, "1": 0.52, "2": 0.15, "3": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:c77d70a98fccfddd4608e441", "state_id": "grid_navigation:c77d70a98fccfddd4608e441", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG # . .\n. # . .\n. . A .\n# # # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.0, "west": 1.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (1, 1), (3, 0), (3, 1), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 1], [3, 0], [3, 1], [3, 2], [3, 3]], "position": [2, 2], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.12, "north": 0.63, "east": 0.25}, "solvable": {"false": 0.49, "true": 0.51}, "value": {"0": 0.04, "1": 0.47, "2": 0.29, "3": 0.2}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:881801f80336e899241200e9", "state_id": "grid_navigation:881801f80336e899241200e9", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\nG . # A\n. . . .\n. . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.5, "south": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((2, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 2], [3, 3]], "position": [1, 3], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.21, "south": 0.79}, "solvable": {"false": 0.31000000000000005, "true": 0.69}, "value": {"0": 0.08, "1": 0.58, "2": 0.28, "3": 0.06}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:afa7b02dd8e8e5f75566c7bb", "state_id": "tic_tac_toe:afa7b02dd8e8e5f75566c7bb", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O .\nX . X\nO X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place O in row 1, column 3 (cell 3).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_1": "Place O in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_3": 0.0, "cell_9": 0.0, "cell_5": 1.0, "cell_1": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.O.X.X.XO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".O.X.XOX.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.13, "cell_9": 0.37, "cell_3": 0.15, "cell_5": 0.35}, "solvable": {"false": 0.86, "true": 0.14}, "value": {"0": 0.37, "1": 0.41, "2": 0.22}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:557b009cc31558f5c324f37e", "state_id": "grid_navigation:557b009cc31558f5c324f37e", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # A .\nG # . .\n# # . #\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"south": 0.5, "east": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 0), (1, 1), (1, 2), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 1], [2, 0], [2, 1], [2, 3]], "position": [0, 2], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.5, "east": 0.5}, "solvable": {"false": 0.49, "true": 0.51}, "value": {"0": 0.03, "1": 0.29, "2": 0.48, "3": 0.2}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:5757b9776932a73cec93a328", "state_id": "grid_navigation:5757b9776932a73cec93a328", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . A .\n. . . G\n. . . .\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.5, "west": 0.0, "south": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=()", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [], "position": [0, 2], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0, "south": 0.14, "east": 0.86}, "solvable": {"false": 0.030000000000000027, "true": 0.97}, "value": {"0": 0.52, "1": 0.48, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:2d1d0930528e4a102aff3d6d", "state_id": "grid_navigation:2d1d0930528e4a102aff3d6d", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . A #\n# . . .\n. # # #\n# # G .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.5, "south": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (0, 3), (1, 0), (1, 1), (1, 2), (2, 3), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 0], [2, 1], [2, 2], [2, 3], [3, 0], [3, 1]], "position": [0, 2], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.42, "south": 0.58}, "solvable": {"false": 0.5800000000000001, "true": 0.42}, "value": {"0": 0.01, "1": 0.33, "2": 0.3, "3": 0.36}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:42a285c7f48aee4f47849e12", "state_id": "grid_navigation:42a285c7f48aee4f47849e12", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\n. . . #\nA G # .\n. . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 1.0, "north": 0.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 2), (0, 3), (1, 2), (2, 3), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 3], [2, 2], [3, 2], [3, 3]], "position": [2, 0], "goal": [2, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.05, "east": 0.67, "north": 0.27}, "solvable": {"false": 0.040000000000000036, "true": 0.96}, "value": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:5731f4c66cc238225a8ff21a", "state_id": "grid_navigation:5731f4c66cc238225a8ff21a", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG . . .\nA . . .\n. . . #\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.0, "north": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((2, 3),)", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[2, 3]], "position": [1, 0], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.01, "east": 0.13, "north": 0.86}, "solvable": {"false": 0.08999999999999997, "true": 0.91}, "value": {"0": 0.83, "1": 0.15, "2": 0.02, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:c848dc32ff369fb846154459", "state_id": "grid_navigation:c848dc32ff369fb846154459", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . A\n. . . G\n. . . .\n. # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 1.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((2, 3),)", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[3, 1]], "position": [0, 3], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.04, "south": 0.96}, "solvable": {"false": 0.06000000000000005, "true": 0.94}, "value": {"0": 0.98, "1": 0.02, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:5a472b2e979fcec1f75d0c23", "state_id": "grid_navigation:5a472b2e979fcec1f75d0c23", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . G\nA # . #\n. . . .\n. # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 2}, "gold_probs": {"action": {"north": 0.0, "south": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (2, 0), (2, 1), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 1], [1, 3], [3, 1]], "position": [1, 0], "goal": [0, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.72, "north": 0.28}, "solvable": {"false": 0.42000000000000004, "true": 0.58}, "value": {"0": 0.02, "1": 0.41, "2": 0.39, "3": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:03f3bfa7639a2cd9078c6030", "state_id": "grid_navigation:03f3bfa7639a2cd9078c6030", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . G .\n# . # .\n. . . .\n. # . A", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 1.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 1), (1, 3), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [1, 2], [3, 1]], "position": [3, 3], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.3, "north": 0.7}, "solvable": {"false": 0.17000000000000004, "true": 0.83}, "value": {"0": 0.01, "1": 0.44, "2": 0.48, "3": 0.07}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:1e70a8391f3b5f4df313f1cf", "state_id": "grid_navigation:1e70a8391f3b5f4df313f1cf", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nA . . #\n. # # .\n# . . .\n. . G .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "south": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 3), (2, 1), (2, 2), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 1], [1, 2], [2, 0]], "position": [0, 0], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.16, "south": 0.84}, "solvable": {"false": 0.27, "true": 0.73}, "value": {"0": 0.01, "1": 0.44, "2": 0.45, "3": 0.1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:2a2a1bfd8c6d2b676e47f392", "state_id": "tic_tac_toe:2a2a1bfd8c6d2b676e47f392", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\n. O X\nX O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.0, "cell_2": 1.0, "cell_4": 0.0, "cell_3": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....OXXOX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "....OXXOX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.38, "cell_3": 0.21, "cell_2": 0.18, "cell_1": 0.23}, "solvable": {"false": 0.88, "true": 0.12}, "value": {"0": 0.4, "1": 0.45, "2": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:1f97b33aedebbc8f5cb6f2d4", "state_id": "tic_tac_toe:1f97b33aedebbc8f5cb6f2d4", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO O .\n. . X\nO X X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place X in row 2, column 2 (cell 5).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_4": "Place X in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 0.0, "cell_3": 1.0, "cell_4": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOX..XXO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OO...XOXX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.18, "cell_3": 0.48, "cell_5": 0.34}, "solvable": {"false": 0.54, "true": 0.46}, "value": {"0": 0.21, "1": 0.29, "2": 0.5}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:838c705c61123ce141b3cbb3", "state_id": "grid_navigation:838c705c61123ce141b3cbb3", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A . #\n. # . .\nG . . #\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 1.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 2), (3, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 1], [2, 3]], "position": [0, 1], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.04, "west": 0.96}, "solvable": {"false": 0.19999999999999996, "true": 0.8}, "value": {"0": 0.05, "1": 0.7, "2": 0.21, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:439402bf7b2bc1711c360155", "state_id": "tic_tac_toe:439402bf7b2bc1711c360155", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O .\nX X O\nX O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place X in row 1, column 3 (cell 3).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_1": "Place X in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_3": 0.3333333333333333, "cell_9": 0.3333333333333333, "cell_1": 0.3333333333333333}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_3", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.O.OXO.XX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".O.XXOXO.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.16, "cell_9": 0.56, "cell_1": 0.27}, "solvable": {"false": 0.4, "true": 0.6}, "value": {"0": 0.09, "1": 0.24, "2": 0.67}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:91c36d816324fda6d7e5ceb2", "state_id": "tic_tac_toe:91c36d816324fda6d7e5ceb2", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . O\nX O .\nX . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.3333333333333333, "cell_2": 0.0, "cell_9": 0.3333333333333333, "cell_8": 0.3333333333333333, "cell_6": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O.O.XX.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..OXO.X..", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.16, "cell_6": 0.24, "cell_9": 0.18, "cell_2": 0.16, "cell_8": 0.26}, "solvable": {"false": 0.6799999999999999, "true": 0.32}, "value": {"0": 0.2, "1": 0.3, "2": 0.5}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:8616daf22b59cf1bea2c0afa", "state_id": "grid_navigation:8616daf22b59cf1bea2c0afa", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG . A .\n# . . .\n. . . .\n# . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "east": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [3, 0]], "position": [0, 2], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.09, "south": 0.11, "west": 0.8}, "solvable": {"false": 0.18999999999999995, "true": 0.81}, "value": {"0": 0.85, "1": 0.11, "2": 0.03, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:719fd506bed0b71df6df93d9", "state_id": "tic_tac_toe:719fd506bed0b71df6df93d9", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\nO X .\nX X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place O in row 2, column 3 (cell 6).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_1": "Place O in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_6": 0.25, "cell_2": 0.25, "cell_3": 0.25, "cell_1": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_2", "cell_3", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....XOOXX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "...OX.XXO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.26, "cell_3": 0.28, "cell_6": 0.3, "cell_2": 0.16}, "solvable": {"false": 0.86, "true": 0.14}, "value": {"0": 0.62, "1": 0.19, "2": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:85751721d72ebef2705a6eaf", "state_id": "tic_tac_toe:85751721d72ebef2705a6eaf", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . .\n. O X\n. X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place O in row 1, column 2 (cell 2).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_2": 0.25, "cell_4": 0.25, "cell_7": 0.25, "cell_3": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_3", "cell_4", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XXO.OX.:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X...OX.XO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.29, "cell_4": 0.3, "cell_3": 0.19, "cell_2": 0.22}, "solvable": {"false": 0.85, "true": 0.15}, "value": {"0": 0.33, "1": 0.42, "2": 0.25}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:e37417a2b05af3a70de27a09", "state_id": "grid_navigation:e37417a2b05af3a70de27a09", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # .\n# G . .\n. . # .\nA . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.5, "east": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 1), (2, 0), (2, 2), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [1, 0], [2, 2], [3, 2]], "position": [3, 0], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.82, "east": 0.18}, "solvable": {"false": 0.19999999999999996, "true": 0.8}, "value": {"0": 0.1, "1": 0.6, "2": 0.21, "3": 0.09}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:335c00c1afb8fbcb00ef5283", "state_id": "grid_navigation:335c00c1afb8fbcb00ef5283", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# G # .\n. # . A\n. . . #\n# . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 2), (1, 1), (2, 3), (3, 0), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [1, 1], [2, 3], [3, 0], [3, 2]], "position": [1, 3], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.4, "west": 0.6}, "solvable": {"false": 0.63, "true": 0.37}, "value": {"0": 0.03, "1": 0.31, "2": 0.32, "3": 0.33999999999999997}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:8e7a4b5d6caacf2cd4d50c07", "state_id": "tic_tac_toe:8e7a4b5d6caacf2cd4d50c07", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . X\nO O X\nX O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place X in row 1, column 2 (cell 2).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_1": "Place X in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.5, "cell_9": 0.5, "cell_1": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOOXXO.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..XOOXXO.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.75, "cell_2": 0.08, "cell_1": 0.17}, "solvable": {"false": 0.41000000000000003, "true": 0.59}, "value": {"0": 0.09, "1": 0.22, "2": 0.69}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:474d1ab0c9fe12b91648f218", "state_id": "tic_tac_toe:474d1ab0c9fe12b91648f218", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X O\nX O X\n. O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_7", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_1": 0.0, "cell_9": 0.0, "cell_7": 1.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.O.XOX.XO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XOXOX.O.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.4, "cell_9": 0.27, "cell_1": 0.33}, "solvable": {"false": 0.6699999999999999, "true": 0.33}, "value": {"0": 0.32, "1": 0.42, "2": 0.26}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:a8412148741bf8a9a7639449", "state_id": "grid_navigation:a8412148741bf8a9a7639449", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . .\n# . . .\n. . . .\nA . . G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((2, 3), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 0]], "position": [3, 0], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.08, "east": 0.92}, "solvable": {"false": 0.06000000000000005, "true": 0.94}, "value": {"0": 0.09, "1": 0.77, "2": 0.13, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:4e75bcdc942b9986c4604560", "state_id": "grid_navigation:4e75bcdc942b9986c4604560", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n. . . #\n# # A .\nG # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.3333333333333333, "south": 0.3333333333333333, "east": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (1, 0), (1, 1), (2, 3), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 3], [2, 0], [2, 1], [3, 1]], "position": [2, 2], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.32, "east": 0.11, "south": 0.56}, "solvable": {"false": 0.43000000000000005, "true": 0.57}, "value": {"0": 0.04, "1": 0.44, "2": 0.35, "3": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:95ea6fff1237dbd0c872cce2", "state_id": "grid_navigation:95ea6fff1237dbd0c872cce2", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # # #\n. # . .\n. . G #\n# . A .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 1.0, "east": 0.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (0, 3), (1, 3), (2, 2), (2, 3), (3, 0), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [0, 2], [0, 3], [1, 1], [2, 3], [3, 0]], "position": [3, 2], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.29, "north": 0.6, "east": 0.11}, "solvable": {"false": 0.17000000000000004, "true": 0.83}, "value": {"0": 0.56, "1": 0.35, "2": 0.05, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:7bdc37c0c6017ca7d08d8f88", "state_id": "grid_navigation:7bdc37c0c6017ca7d08d8f88", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . #\n# # . .\nA . # .\n. G # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.5, "east": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 2), (2, 0), (2, 1), (3, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 3], [1, 0], [1, 1], [2, 2], [3, 2]], "position": [2, 0], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.61, "east": 0.39}, "solvable": {"false": 0.20999999999999996, "true": 0.79}, "value": {"0": 0.8, "1": 0.1, "2": 0.05, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:9ff306524f4691d6b59f9be5", "state_id": "grid_navigation:9ff306524f4691d6b59f9be5", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # #\n# # G .\n# # # .\n# . A .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 1.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 1), (1, 2), (1, 3), (2, 1), (2, 2), (2, 3), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [1, 0], [1, 1], [2, 0], [2, 1], [2, 2], [3, 0]], "position": [3, 2], "goal": [1, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.64, "west": 0.36}, "solvable": {"false": 0.62, "true": 0.38}, "value": {"0": 0.04, "1": 0.29, "2": 0.26, "3": 0.41}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:9a7d7c5ef66554b5673d3e88", "state_id": "grid_navigation:9a7d7c5ef66554b5673d3e88", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # G .\n# . # #\n. . . .\n. . . A", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.5, "west": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 0), (1, 1), (1, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 0], [1, 2], [1, 3]], "position": [3, 3], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.56, "west": 0.44}, "solvable": {"false": 0.41000000000000003, "true": 0.59}, "value": {"0": 0, "1": 0.21, "2": 0.63, "3": 0.16}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:3bd364e750c99bfdb1b390f7", "state_id": "tic_tac_toe:3bd364e750c99bfdb1b390f7", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X X\n. . .\nO X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place X in row 2, column 3 (cell 6).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_4": "Place X in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.0, "cell_5": 1.0, "cell_4": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.OX.XO.X:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXX...OXO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.16, "cell_4": 0.32, "cell_5": 0.52}, "solvable": {"false": 0.49, "true": 0.51}, "value": {"0": 0.08, "1": 0.33, "2": 0.59}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:dd8d4ac3c52cda6e45f3ba72", "state_id": "tic_tac_toe:dd8d4ac3c52cda6e45f3ba72", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X O\nX X .\nX . O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place O in row 2, column 3 (cell 6).", "cell_8": "Place O in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_6", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 1.0, "cell_8": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.O.XXXXO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXOXX.X.O", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.49, "cell_8": 0.51}, "solvable": {"false": 0.85, "true": 0.15}, "value": {"0": 0.67, "1": 0.19, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:db0e0da814811e8508ef50f9", "state_id": "tic_tac_toe:db0e0da814811e8508ef50f9", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X O\nO . X\nX O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_9": 0.3333333333333333, "cell_1": 0.3333333333333333, "cell_5": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_5", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXX.OOX.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XOO.XXO.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.39999999999999997, "cell_9": 0.4, "cell_1": 0.2}, "solvable": {"false": 0.6599999999999999, "true": 0.34}, "value": {"0": 0.1, "1": 0.45999999999999996, "2": 0.44}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:0306b7412defc96e615c6968", "state_id": "grid_navigation:0306b7412defc96e615c6968", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\n. # A #\n# . . G\n. # . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 1.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 2), (2, 0), (2, 2), (2, 3), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 1], [1, 3], [2, 0], [3, 1], [3, 3]], "position": [1, 2], "goal": [2, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.82, "north": 0.18}, "solvable": {"false": 0.28, "true": 0.72}, "value": {"0": 0.44, "1": 0.42, "2": 0.06, "3": 0.08}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:cbf3ac462afcbd5c9a9f2c4e", "state_id": "grid_navigation:cbf3ac462afcbd5c9a9f2c4e", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # .\n# . # A\n. # # .\nG # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.5, "south": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (1, 1), (1, 2), (2, 0), (2, 2), (3, 0), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [1, 0], [1, 2], [2, 1], [2, 2], [3, 1]], "position": [1, 3], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.34, "south": 0.66}, "solvable": {"false": 0.53, "true": 0.47}, "value": {"0": 0.01, "1": 0.31, "2": 0.38, "3": 0.3}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:b64650a7fc48cfcdf5a22f2c", "state_id": "grid_navigation:b64650a7fc48cfcdf5a22f2c", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . .\n. # A #\n. # . #\nG # . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.5, "south": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 3), (1, 1), (1, 3), (2, 1), (2, 3), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 1], [1, 3], [2, 1], [2, 3], [3, 1], [3, 3]], "position": [1, 2], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.15, "south": 0.85}, "solvable": {"false": 0.53, "true": 0.47}, "value": {"0": 0.03, "1": 0.45, "2": 0.26, "3": 0.26}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:d33fe66d61f5ad820f111ebb", "state_id": "grid_navigation:d33fe66d61f5ad820f111ebb", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A . .\n# . # #\n. # # .\n. # # G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.3333333333333333, "west": 0.3333333333333333, "south": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 2), (1, 1), (1, 2), (2, 0), (2, 1), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [1, 2], [1, 3], [2, 1], [2, 2], [3, 1], [3, 2]], "position": [0, 1], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.36, "west": 0.09, "east": 0.55}, "solvable": {"false": 0.7, "true": 0.3}, "value": {"0": 0.01, "1": 0.22, "2": 0.32, "3": 0.45}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:3f9ac5d507c0766ea0cb56b4", "state_id": "grid_navigation:3f9ac5d507c0766ea0cb56b4", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # .\n. G # .\n. A . .\n. . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east.", "south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 1.0, "east": 0.0, "south": 0.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (0, 2), (1, 2), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [1, 2], [3, 2], [3, 3]], "position": [2, 1], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.08, "north": 0.87, "south": 0.01, "east": 0.04}, "solvable": {"false": 0.31999999999999995, "true": 0.68}, "value": {"0": 0.55, "1": 0.28, "2": 0.09, "3": 0.08}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:1cb597ecec47f264501ce007", "state_id": "tic_tac_toe:1cb597ecec47f264501ce007", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO O X\nX . .\nX . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.0, "cell_8": 0.0, "cell_6": 0.0, "cell_5": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..X..OXXO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OOXX..X..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.1, "cell_5": 0.6, "cell_6": 0.15, "cell_8": 0.15}, "solvable": {"false": 0.85, "true": 0.15}, "value": {"0": 0.75, "1": 0.15, "2": 0.1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:6c253d89221a9644b4091cf7", "state_id": "grid_navigation:6c253d89221a9644b4091cf7", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # G\n. # . #\n. . . .\n. # . A", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.5, "west": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (1, 0), (1, 2), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 1], [1, 3], [3, 1]], "position": [3, 3], "goal": [0, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.66, "west": 0.34}, "solvable": {"false": 0.31999999999999995, "true": 0.68}, "value": {"0": 0.01, "1": 0.3, "2": 0.54, "3": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:8f75510d6c0a251474b9e024", "state_id": "tic_tac_toe:8f75510d6c0a251474b9e024", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X .\n. X O\nX . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_4": "Place O in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_9": 0.25, "cell_3": 0.25, "cell_8": 0.25, "cell_4": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_4", "cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOX..XO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OX..XOX..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.24, "cell_9": 0.24, "cell_3": 0.18, "cell_8": 0.34}, "solvable": {"false": 0.88, "true": 0.12}, "value": {"0": 0.66, "1": 0.19, "2": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:9734cba62c0733e7cc52fb22", "state_id": "tic_tac_toe:9734cba62c0733e7cc52fb22", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O O\n. . X\nX X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place O in row 2, column 1 (cell 4).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_9": 0.0, "cell_1": 1.0, "cell_5": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XO.XOX.:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".OO..XXX.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.22, "cell_5": 0.2, "cell_9": 0.43, "cell_4": 0.15}, "solvable": {"false": 0.79, "true": 0.21}, "value": {"0": 0.49, "1": 0.27, "2": 0.24}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:d9dea8c687370a236b9b3c57", "state_id": "grid_navigation:d9dea8c687370a236b9b3c57", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\nG # . A\n. . # .\n# # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.0, "north": 0.5, "west": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 1), (1, 3), (2, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 1], [2, 2], [3, 0], [3, 1]], "position": [1, 3], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.58, "north": 0.15, "south": 0.27}, "solvable": {"false": 0.5700000000000001, "true": 0.43}, "value": {"0": 0.08, "1": 0.32, "2": 0.27, "3": 0.33}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:88f3d24439a45dbf1699b1b8", "state_id": "grid_navigation:88f3d24439a45dbf1699b1b8", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. G # .\n. . . #\n. # A .\n# . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 1.0, "east": 0.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 3), (2, 1), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 3], [2, 1], [3, 0]], "position": [2, 2], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.25, "south": 0.2, "north": 0.55}, "solvable": {"false": 0.28, "true": 0.72}, "value": {"0": 0.02, "1": 0.46, "2": 0.37, "3": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:9dfa2bb63e0aeefb17094176", "state_id": "grid_navigation:9dfa2bb63e0aeefb17094176", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\nA G . #\n# . # .\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 1.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 2), (1, 3), (2, 0), (2, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 3], [2, 0], [2, 2]], "position": [1, 0], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.22, "east": 0.78}, "solvable": {"false": 0.040000000000000036, "true": 0.96}, "value": {"0": 0.98, "1": 0.01, "2": 0, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:a76ff33fdb44e052b75ec57e", "state_id": "grid_navigation:a76ff33fdb44e052b75ec57e", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . .\n. # A .\n. . . G\n. . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "east": 0.5, "south": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (2, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 1], [3, 3]], "position": [1, 2], "goal": [2, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.03, "south": 0.53, "east": 0.44}, "solvable": {"false": 0.12, "true": 0.88}, "value": {"0": 0.72, "1": 0.24, "2": 0.02, "3": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:009d13e8466e0ea7bae02db5", "state_id": "grid_navigation:009d13e8466e0ea7bae02db5", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\n. G # .\n# # . .\n. . . A", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.5, "west": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (1, 2), (2, 0), (2, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 2], [2, 0], [2, 1]], "position": [3, 3], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.48, "north": 0.52}, "solvable": {"false": 0.42000000000000004, "true": 0.58}, "value": {"0": 0.01, "1": 0.37, "2": 0.43, "3": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:2f16a806d64ede870b310e25", "state_id": "tic_tac_toe:2f16a806d64ede870b310e25", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X O\n. X X\n. O O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place O in row 2, column 1 (cell 4).", "cell_7": "Place O in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_7", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_7": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOXXOXO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XXO.XX.OO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.51, "cell_7": 0.49}, "solvable": {"false": 0.79, "true": 0.21}, "value": {"0": 0.47, "1": 0.32, "2": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:a218a5b0f4f7db6b7d43e7fe", "state_id": "grid_navigation:a218a5b0f4f7db6b7d43e7fe", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG . . #\n. . # #\n# . . .\n. A # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (2, 1), (2, 3), (3, 0), (3, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 2], [1, 3], [2, 0], [3, 2], [3, 3]], "position": [3, 1], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.32, "north": 0.68}, "solvable": {"false": 0.41000000000000003, "true": 0.59}, "value": {"0": 0.03, "1": 0.38, "2": 0.37, "3": 0.22}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:374070a39c2e22527a04f950", "state_id": "grid_navigation:374070a39c2e22527a04f950", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# A . .\n. . # G\n# . . .\n. # . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 1), (2, 3), (3, 0), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 2], [2, 0], [3, 1], [3, 3]], "position": [0, 1], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.85, "south": 0.15}, "solvable": {"false": 0.22999999999999998, "true": 0.77}, "value": {"0": 0.06, "1": 0.68, "2": 0.17, "3": 0.09}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:93dd0f6d69db0144d2106e13", "state_id": "grid_navigation:93dd0f6d69db0144d2106e13", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. . # .\n# A # .\n# . # G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 2}, "gold_probs": {"action": {"south": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 3), (1, 1), (1, 3), (2, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 2], [2, 0], [2, 2], [3, 0], [3, 2]], "position": [2, 1], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.5, "south": 0.5}, "solvable": {"false": 0.5700000000000001, "true": 0.43}, "value": {"0": 0.04, "1": 0.33999999999999997, "2": 0.33, "3": 0.29}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:f569bbb2ac70d48b321a637c", "state_id": "tic_tac_toe:f569bbb2ac70d48b321a637c", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X O\n. . X\n. . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place X in row 2, column 2 (cell 5).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_4": "Place X in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 1.0, "cell_8": 0.0, "cell_9": 0.0, "cell_7": 0.0, "cell_4": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.....XOXO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXO..X...", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.16, "cell_9": 0.15, "cell_5": 0.54, "cell_7": 0.09, "cell_8": 0.06}, "solvable": {"false": 0.8200000000000001, "true": 0.18}, "value": {"0": 0.22, "1": 0.49, "2": 0.29}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:675fa54fc71b14adbc9be0e4", "state_id": "grid_navigation:675fa54fc71b14adbc9be0e4", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n# . . .\n. G # .\n. A . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 1.0, "west": 0.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 2), (2, 1), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 0], [2, 2]], "position": [3, 1], "goal": [2, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.05, "north": 0.86, "west": 0.09}, "solvable": {"false": 0.21999999999999997, "true": 0.78}, "value": {"0": 0.52, "1": 0.34, "2": 0.09, "3": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:76aaee4f12e6335b0f3e2f55", "state_id": "tic_tac_toe:76aaee4f12e6335b0f3e2f55", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X .\nO . X\nO X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place X in row 1, column 3 (cell 3).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_3": 0.0, "cell_1": 0.5, "cell_5": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOX.X.XO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".X.O.XOXO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.48, "cell_3": 0.23, "cell_1": 0.29}, "solvable": {"false": 0.6, "true": 0.4}, "value": {"0": 0.16, "1": 0.39, "2": 0.45}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:756eb7631fab609f1c08f629", "state_id": "grid_navigation:756eb7631fab609f1c08f629", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . .\n. # # #\n. . . .\n. . A G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "east": 1.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (1, 2), (2, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 1], [1, 2], [1, 3]], "position": [3, 2], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.97, "west": 0.01, "north": 0.02}, "solvable": {"false": 0.06000000000000005, "true": 0.94}, "value": {"0": 0.98, "1": 0.01, "2": 0, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:45912987dc92378166098a2e", "state_id": "tic_tac_toe:45912987dc92378166098a2e", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\nX O X\nO X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place X in row 1, column 2 (cell 2).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_3": "Place X in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_2": 0.3333333333333333, "cell_1": 0.3333333333333333, "cell_3": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_2", "cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...XOXOXO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "...XOXOXO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.16, "cell_2": 0.21, "cell_1": 0.63}, "solvable": {"false": 0.72, "true": 0.28}, "value": {"0": 0.08, "1": 0.52, "2": 0.4}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:5f8fe5669ddc868486e4724a", "state_id": "grid_navigation:5f8fe5669ddc868486e4724a", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . A\nG . # .\n. # # .\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 2), (2, 1), (2, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 2], [2, 1], [2, 2]], "position": [0, 3], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.11, "west": 0.89}, "solvable": {"false": 0.36, "true": 0.64}, "value": {"0": 0.07, "1": 0.5, "2": 0.32, "3": 0.11}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:ad215b8f5c13885a0d7d34fd", "state_id": "tic_tac_toe:ad215b8f5c13885a0d7d34fd", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O X\n. . O\nX . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place O in row 2, column 2 (cell 5).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_8": "Place O in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_5": 0.25, "cell_4": 0.25, "cell_9": 0.25, "cell_8": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4", "cell_5", "cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XO..XOX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOX..OX..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.12, "cell_4": 0.19, "cell_8": 0.21, "cell_5": 0.48}, "solvable": {"false": 0.86, "true": 0.14}, "value": {"0": 0.48, "1": 0.32, "2": 0.2}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:dfff93a377a0676735ebf6aa", "state_id": "tic_tac_toe:dfff93a377a0676735ebf6aa", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . X\nX O .\n. O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place O in row 2, column 3 (cell 6).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_1": "Place O in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.0, "cell_7": 0.0, "cell_2": 1.0, "cell_1": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XXO..OX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..XXO..OX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.32, "cell_7": 0.18, "cell_2": 0.25, "cell_1": 0.25}, "solvable": {"false": 0.9, "true": 0.1}, "value": {"0": 0.42, "1": 0.43, "2": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:f66b541cf66219744b31b3e2", "state_id": "grid_navigation:f66b541cf66219744b31b3e2", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nA . . .\n. . . #\n. . . #\n. G . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.5, "east": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 3), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 3], [2, 3], [3, 3]], "position": [0, 0], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.06, "south": 0.94}, "solvable": {"false": 0.07999999999999996, "true": 0.92}, "value": {"0": 0.04, "1": 0.66, "2": 0.26, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:d0dafcb91f30976024ff8871", "state_id": "tic_tac_toe:d0dafcb91f30976024ff8871", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X O\nX . X\n. . O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place X in row 2, column 2 (cell 5).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_8": "Place X in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 1.0, "cell_7": 0.0, "cell_8": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OX.XOXO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXOX.X..O", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.18, "cell_7": 0.22, "cell_5": 0.6}, "solvable": {"false": 0.53, "true": 0.47}, "value": {"0": 0.14, "1": 0.32, "2": 0.54}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:766e777a55f55b397e7e873c", "state_id": "grid_navigation:766e777a55f55b397e7e873c", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. . G .\n. # # A\n# . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((1, 2), (2, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[2, 1], [2, 2], [3, 0]], "position": [2, 3], "goal": [1, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.81, "south": 0.19}, "solvable": {"false": 0.31999999999999995, "true": 0.68}, "value": {"0": 0.24, "1": 0.54, "2": 0.16, "3": 0.06}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:42caa547cb595cb440e6929f", "state_id": "grid_navigation:42caa547cb595cb440e6929f", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A . .\n. . . .\n# # . .\n# . G #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 0.0, "east": 0.5, "south": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 3), (1, 2), (1, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[2, 0], [2, 1], [3, 0], [3, 3]], "position": [0, 1], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.05, "south": 0.61, "east": 0.34}, "solvable": {"false": 0.09999999999999998, "true": 0.9}, "value": {"0": 0, "1": 0.38, "2": 0.56, "3": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:5bdb824988a155ad30b503c0", "state_id": "grid_navigation:5bdb824988a155ad30b503c0", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # #\n. . . G\n# . . #\n# A . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.5, "east": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 2), (0, 3), (1, 0), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [2, 0], [2, 3], [3, 0], [3, 3]], "position": [3, 1], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.55, "east": 0.45}, "solvable": {"false": 0.09999999999999998, "true": 0.9}, "value": {"0": 0.07, "1": 0.78, "2": 0.12, "3": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:f4e2d94d687b668ec172764e", "state_id": "tic_tac_toe:f4e2d94d687b668ec172764e", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X .\n. . .\nO O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_6": 0.0, "cell_9": 0.5, "cell_3": 0.5, "cell_5": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...O.XO.X:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XX....OO.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.47, "cell_9": 0.2, "cell_5": 0.19, "cell_6": 0.07, "cell_4": 0.07}, "solvable": {"false": 0.45999999999999996, "true": 0.54}, "value": {"0": 0.14, "1": 0.28, "2": 0.58}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:04a7936cf23e7d180499b310", "state_id": "tic_tac_toe:04a7936cf23e7d180499b310", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O .\nO . X\n. . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place X in row 3, column 2 (cell 8).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_8": 0.0, "cell_7": 0.0, "cell_3": 0.0, "cell_5": 0.0, "cell_9": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...O.XXO.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XO.O.X...", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.1, "cell_5": 0.56, "cell_9": 0.12, "cell_7": 0.1, "cell_3": 0.12}, "solvable": {"false": 0.8200000000000001, "true": 0.18}, "value": {"0": 0.2, "1": 0.54, "2": 0.26}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:261f5a94c276a513337e7911", "state_id": "grid_navigation:261f5a94c276a513337e7911", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. . . A\n. # # #\n# . . G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.5, "west": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (1, 1), (2, 1), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[2, 1], [2, 2], [2, 3], [3, 0]], "position": [1, 3], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.25, "west": 0.75}, "solvable": {"false": 0.52, "true": 0.48}, "value": {"0": 0.02, "1": 0.34, "2": 0.33, "3": 0.31}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:226a16b9803fe9bdf2b77a39", "state_id": "grid_navigation:226a16b9803fe9bdf2b77a39", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . A .\n. . . .\n. # # .\n. . . G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.5, "west": 0.0, "east": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((1, 1), (1, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [2, 1], [2, 2]], "position": [0, 2], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.29, "south": 0.6, "west": 0.11}, "solvable": {"false": 0.15000000000000002, "true": 0.85}, "value": {"0": 0, "1": 0.44, "2": 0.5, "3": 0.06}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:c0fe26ea53c12fb70917c80a", "state_id": "tic_tac_toe:c0fe26ea53c12fb70917c80a", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . O\nX O .\nX X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place O in row 2, column 3 (cell 6).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_1": "Place O in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_6": 0.25, "cell_2": 0.25, "cell_9": 0.25, "cell_1": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_2", "cell_6", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXO.XX.:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..OXO.XX.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.16, "cell_2": 0.08, "cell_6": 0.29, "cell_9": 0.47}, "solvable": {"false": 0.81, "true": 0.19}, "value": {"0": 0.4, "1": 0.28, "2": 0.32}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:585256266af043df95703a12", "state_id": "grid_navigation:585256266af043df95703a12", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # .\n# . . #\n. . A .\n# . G .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north.", "west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 1.0, "north": 0.0, "west": 0.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (2, 0), (2, 3), (3, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [1, 0], [1, 3], [3, 0]], "position": [2, 2], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.3, "east": 0.01, "south": 0.64, "north": 0.05}, "solvable": {"false": 0.09999999999999998, "true": 0.9}, "value": {"0": 0.7, "1": 0.25, "2": 0.03, "3": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:093ad0683b0118306f86d7a0", "state_id": "tic_tac_toe:093ad0683b0118306f86d7a0", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . O\n. X X\nO X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_1": 0.25, "cell_4": 0.25, "cell_9": 0.25, "cell_2": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_2", "cell_4", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O.XXOX.:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..O.XXOX.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.45, "cell_2": 0.14, "cell_1": 0.14, "cell_4": 0.27}, "solvable": {"false": 0.83, "true": 0.17}, "value": {"0": 0.41000000000000003, "1": 0.38, "2": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:b01bb6b910c1d1d0fae4ed50", "state_id": "tic_tac_toe:b01bb6b910c1d1d0fae4ed50", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X .\n. O X\nO X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place O in row 1, column 3 (cell 3).", "cell_4": "Place O in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_3": 1.0, "cell_4": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.XOXOXX.O:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XX..OXOXO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.6, "cell_4": 0.4}, "solvable": {"false": 0.84, "true": 0.16}, "value": {"0": 0.38, "1": 0.42, "2": 0.2}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:7fc070a90968feff75d9364e", "state_id": "tic_tac_toe:7fc070a90968feff75d9364e", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X O\n. . .\n. X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_1": "Place X in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_5": 1.0, "cell_7": 0.0, "cell_6": 0.0, "cell_1": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...X.XO.O:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XO....XO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.05, "cell_7": 0.13, "cell_4": 0.16, "cell_5": 0.46, "cell_1": 0.2}, "solvable": {"false": 0.79, "true": 0.21}, "value": {"0": 0.31, "1": 0.46, "2": 0.23}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:04b1da21482af4a13e57f7ef", "state_id": "grid_navigation:04b1da21482af4a13e57f7ef", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . .\n. A # .\n. . . G\n. . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.0, "south": 1.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (1, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 2], [3, 3]], "position": [1, 1], "goal": [2, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.12, "north": 0.06, "south": 0.82}, "solvable": {"false": 0.13, "true": 0.87}, "value": {"0": 0.2, "1": 0.68, "2": 0.08, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:bcff99ab0919da14822706e6", "state_id": "grid_navigation:bcff99ab0919da14822706e6", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n# # # G\n. . # .\n. A . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.3333333333333333, "west": 0.3333333333333333, "north": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 1), (1, 2), (2, 1), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [1, 1], [1, 2], [2, 2], [3, 3]], "position": [3, 1], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.14, "north": 0.64, "east": 0.22}, "solvable": {"false": 0.47, "true": 0.53}, "value": {"0": 0.02, "1": 0.44, "2": 0.31, "3": 0.23}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:2c0a647a461a6434585a3f7a", "state_id": "grid_navigation:2c0a647a461a6434585a3f7a", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # #\n# . A #\n. . . #\n. G . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.5, "west": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 3), (2, 0), (2, 3), (3, 0), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [0, 3], [1, 0], [1, 3], [2, 3]], "position": [1, 2], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.6, "west": 0.4}, "solvable": {"false": 0.08999999999999997, "true": 0.91}, "value": {"0": 0.09, "1": 0.82, "2": 0.06, "3": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:4eeef526cd880d57f12862f1", "state_id": "grid_navigation:4eeef526cd880d57f12862f1", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nA . # .\n. . . G\n. . # .\n. # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.5, "east": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (1, 2), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [2, 2], [3, 1]], "position": [0, 0], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.8, "east": 0.2}, "solvable": {"false": 0.13, "true": 0.87}, "value": {"0": 0.06, "1": 0.74, "2": 0.15, "3": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:cf705d956d7020132845b5c7", "state_id": "grid_navigation:cf705d956d7020132845b5c7", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n. . G .\n. . # .\n. A # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (2, 1), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [2, 2], [3, 2]], "position": [3, 1], "goal": [1, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.84, "west": 0.16}, "solvable": {"false": 0.31000000000000005, "true": 0.69}, "value": {"0": 0.21, "1": 0.55, "2": 0.16, "3": 0.08}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:1ef217c41847b9cb83cdd0be", "state_id": "tic_tac_toe:1ef217c41847b9cb83cdd0be", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X X\n. O .\n. . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place X in row 3, column 2 (cell 8).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_8": 0.0, "cell_4": 0.0, "cell_7": 0.0, "cell_6": 0.0, "cell_9": 1.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....O.OXX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXX.O....", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.14, "cell_6": 0.18, "cell_9": 0.2, "cell_4": 0.33, "cell_8": 0.15}, "solvable": {"false": 0.6699999999999999, "true": 0.33}, "value": {"0": 0.23, "1": 0.42, "2": 0.35}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:3bc85a3a6a441b0a45a7135b", "state_id": "grid_navigation:3bc85a3a6a441b0a45a7135b", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . #\nA # G .\n. . . #\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.5, "north": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 2), (2, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 1], [2, 3]], "position": [1, 0], "goal": [1, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.73, "north": 0.27}, "solvable": {"false": 0.63, "true": 0.37}, "value": {"0": 0.15, "1": 0.54, "2": 0.2, "3": 0.11}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:c0118453c87f0358f8049e1c", "state_id": "tic_tac_toe:c0118453c87f0358f8049e1c", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O O\n. . .\nX O X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place X in row 2, column 2 (cell 5).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 0.5, "cell_4": 0.5, "cell_6": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4", "cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.XO.OX.X:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOO...XOX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.13, "cell_5": 0.5, "cell_4": 0.37}, "solvable": {"false": 0.71, "true": 0.29}, "value": {"0": 0.34, "1": 0.43, "2": 0.23}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:0efacc35909f10bf1e5031ab", "state_id": "grid_navigation:0efacc35909f10bf1e5031ab", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # G .\n# # A #\n. . . #\n. # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 0), (1, 2), (1, 3), (2, 0), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 0], [1, 1], [1, 3], [2, 3], [3, 1]], "position": [1, 2], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.3, "north": 0.7}, "solvable": {"false": 0.55, "true": 0.45}, "value": {"0": 0.43, "1": 0.18, "2": 0.1, "3": 0.29}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:e0442f6fb38550be87dff0e9", "state_id": "tic_tac_toe:e0442f6fb38550be87dff0e9", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . .\n. . X\nX O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place O in row 1, column 3 (cell 3).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_3": 1.0, "cell_2": 0.0, "cell_4": 0.0, "cell_5": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OX..XOX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O....XXOX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.29, "cell_2": 0.17, "cell_5": 0.37, "cell_3": 0.17}, "solvable": {"false": 0.89, "true": 0.11}, "value": {"0": 0.43, "1": 0.39, "2": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:714301129df961f202b9a4af", "state_id": "tic_tac_toe:714301129df961f202b9a4af", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X X\nX O O\n. . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_8": "Place X in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_9": 1.0, "cell_7": 0.0, "cell_8": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...OOXXXO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXXXOO...", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.15, "cell_7": 0.71, "cell_8": 0.14}, "solvable": {"false": 0.5900000000000001, "true": 0.41}, "value": {"0": 0.16, "1": 0.4, "2": 0.44}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:f6cd87aa7dfcb3b3cabde797", "state_id": "tic_tac_toe:f6cd87aa7dfcb3b3cabde797", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . X\n. O .\nO X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place X in row 1, column 2 (cell 2).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 1.0, "cell_4": 0.0, "cell_6": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.XXO.O.X:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.X.O.OXO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.41, "cell_2": 0.31, "cell_6": 0.28}, "solvable": {"false": 0.72, "true": 0.28}, "value": {"0": 0.21, "1": 0.45, "2": 0.34}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:9063619b181ff876ea3400ac", "state_id": "grid_navigation:9063619b181ff876ea3400ac", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\n. A # G\n# . . .\n. # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 1.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 1), (1, 3), (2, 0), (2, 3), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 2], [2, 0], [3, 1], [3, 2]], "position": [1, 1], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.71, "west": 0.29}, "solvable": {"false": 0.6, "true": 0.4}, "value": {"0": 0.26, "1": 0.26, "2": 0.1, "3": 0.38}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:011d53d3e08b525f819bd05b", "state_id": "grid_navigation:011d53d3e08b525f819bd05b", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n# . A .\n. . # #\n# . . G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (1, 1), (1, 3), (3, 0), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 0], [2, 2], [2, 3], [3, 0]], "position": [1, 2], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.47, "west": 0.53}, "solvable": {"false": 0.29000000000000004, "true": 0.71}, "value": {"0": 0.02, "1": 0.55, "2": 0.31, "3": 0.12}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:e76321603fb2628251e05c20", "state_id": "grid_navigation:e76321603fb2628251e05c20", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # . #\n. G # .\n. A . #\n# # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 1.0, "west": 0.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 1), (0, 3), (1, 2), (2, 3), (3, 0), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [0, 3], [1, 2], [2, 3], [3, 0], [3, 1]], "position": [2, 1], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.87, "west": 0.1, "east": 0.03}, "solvable": {"false": 0.18000000000000005, "true": 0.82}, "value": {"0": 0.83, "1": 0.09, "2": 0.03, "3": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:b706e2a873f33123be3bfcc9", "state_id": "tic_tac_toe:b706e2a873f33123be3bfcc9", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O O\nO . .\n. X X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place X in row 2, column 3 (cell 6).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.3333333333333333, "cell_5": 0.3333333333333333, "cell_7": 0.3333333333333333}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5", "cell_6", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXX.OX.O:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOOO...XX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.25, "cell_5": 0.5, "cell_7": 0.25}, "solvable": {"false": 0.62, "true": 0.38}, "value": {"0": 0.22, "1": 0.36, "2": 0.42}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:931c0342550c4263d4d39742", "state_id": "tic_tac_toe:931c0342550c4263d4d39742", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X X\nX X O\n. O .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_7": "Place O in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_7", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_9": 0.0, "cell_7": 1.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.O.OXXXXO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXXXXO.O.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.26, "cell_7": 0.74}, "solvable": {"false": 0.88, "true": 0.12}, "value": {"0": 0.57, "1": 0.32, "2": 0.11}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:5ddf708f254403a00fa0df76", "state_id": "grid_navigation:5ddf708f254403a00fa0df76", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # .\n. . A .\nG . . .\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.0, "west": 0.5, "south": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 3), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2]], "position": [1, 2], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.22, "west": 0.77, "east": 0.01}, "solvable": {"false": 0.050000000000000044, "true": 0.95}, "value": {"0": 0.11, "1": 0.83, "2": 0.05, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:2a9dde55d910feaae896c8fa", "state_id": "grid_navigation:2a9dde55d910feaae896c8fa", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n# . . G\nA . . #\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [2, 3]], "position": [2, 0], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.96, "south": 0.04}, "solvable": {"false": 0.10999999999999999, "true": 0.89}, "value": {"0": 0.02, "1": 0.7, "2": 0.24, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:60784946646ec7915339e043", "state_id": "grid_navigation:60784946646ec7915339e043", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # . #\n. # . A\n. . . .\n. G . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 0.5, "south": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (2, 1), (3, 0), (3, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [0, 3], [1, 1], [3, 3]], "position": [1, 3], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.71, "west": 0.29}, "solvable": {"false": 0.10999999999999999, "true": 0.89}, "value": {"0": 0.01, "1": 0.74, "2": 0.23, "3": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:83e3829eff69265c5ff0c99a", "state_id": "grid_navigation:83e3829eff69265c5ff0c99a", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # G .\n. # . #\nA . . .\n. # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 1.0, "south": 0.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 0), (1, 2), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 1], [1, 3], [3, 1]], "position": [2, 0], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.41, "east": 0.5599999999999999, "south": 0.03}, "solvable": {"false": 0.48, "true": 0.52}, "value": {"0": 0.01, "1": 0.38, "2": 0.4, "3": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:4d6e3030839729580b2d3e33", "state_id": "grid_navigation:4d6e3030839729580b2d3e33", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. . . .\nA . . #\n# # # G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 2), (0, 3), (1, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[2, 3], [3, 0], [3, 1], [3, 2]], "position": [2, 0], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.17, "east": 0.83}, "solvable": {"false": 0.45999999999999996, "true": 0.54}, "value": {"0": 0.05, "1": 0.43, "2": 0.24, "3": 0.28}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:488a0495307495a633901f43", "state_id": "tic_tac_toe:488a0495307495a633901f43", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . X\nX O O\nX X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 1.0, "cell_2": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XXOOXXO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..XXOOXXO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.51, "cell_2": 0.49}, "solvable": {"false": 0.91, "true": 0.09}, "value": {"0": 0.72, "1": 0.2, "2": 0.08}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:7309349b22e6f0dbfbf69788", "state_id": "grid_navigation:7309349b22e6f0dbfbf69788", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n. . . #\nA . . .\nG . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 1.0, "north": 0.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (2, 3), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 3], [3, 2]], "position": [2, 0], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.05, "north": 0.03, "south": 0.92}, "solvable": {"false": 0.20999999999999996, "true": 0.79}, "value": {"0": 0.77, "1": 0.15, "2": 0.05, "3": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:f4eec2e353ab9191d5a19582", "state_id": "grid_navigation:f4eec2e353ab9191d5a19582", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A . .\n# . . #\nG # . #\n. # # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.3333333333333333, "west": 0.3333333333333333, "south": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 0), (1, 1), (2, 0), (3, 0), (3, 1), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [1, 3], [2, 1], [2, 3], [3, 1], [3, 2], [3, 3]], "position": [0, 1], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.4, "south": 0.36, "east": 0.24}, "solvable": {"false": 0.56, "true": 0.44}, "value": {"0": 0.09, "1": 0.49, "2": 0.17, "3": 0.25}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:44f450e982bb2bfe88670db6", "state_id": "tic_tac_toe:44f450e982bb2bfe88670db6", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O .\nX X .\nO X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_6", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.0, "cell_3": 0.0, "cell_6": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OOXX.XO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".O.XX.OXO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.25, "cell_1": 0.19, "cell_6": 0.56}, "solvable": {"false": 0.47, "true": 0.53}, "value": {"0": 0.1, "1": 0.29, "2": 0.61}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:dcd887c9202322363d51f1b8", "state_id": "tic_tac_toe:dcd887c9202322363d51f1b8", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O .\n. X O\nO X X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place X in row 1, column 3 (cell 3).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_4": "Place X in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_3": 0.0, "cell_1": 1.0, "cell_4": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OOXX.OX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".O..XOOXX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.34, "cell_3": 0.29, "cell_1": 0.37}, "solvable": {"false": 0.5700000000000001, "true": 0.43}, "value": {"0": 0.2, "1": 0.31, "2": 0.49}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:18a91716d1385d0cf2d81657", "state_id": "tic_tac_toe:18a91716d1385d0cf2d81657", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X .\n. O O\nX X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_4": "Place X in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_1": 0.3333333333333333, "cell_3": 0.3333333333333333, "cell_4": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_3", "cell_4"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XXOX.OO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".X..OOXXO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.25, "cell_4": 0.4, "cell_1": 0.35}, "solvable": {"false": 0.5, "true": 0.5}, "value": {"0": 0.16, "1": 0.24, "2": 0.6}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:d61d0f729af26c6caf7a1bae", "state_id": "grid_navigation:d61d0f729af26c6caf7a1bae", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # .\nA . # .\n. # . .\nG . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.0, "south": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 3), (1, 1), (2, 2), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [1, 2], [2, 1]], "position": [1, 0], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.85, "east": 0.15}, "solvable": {"false": 0.26, "true": 0.74}, "value": {"0": 0.07, "1": 0.61, "2": 0.26, "3": 0.06}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:36cf9d1d8490632067206732", "state_id": "grid_navigation:36cf9d1d8490632067206732", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . G A\n# . # .\n# . . .\n# . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 1), (1, 3), (2, 3), (3, 0), (3, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [1, 2], [2, 0], [3, 0], [3, 2], [3, 3]], "position": [0, 3], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.88, "south": 0.12}, "solvable": {"false": 0.22999999999999998, "true": 0.77}, "value": {"0": 0.73, "1": 0.16, "2": 0.06, "3": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:83485c159e2dc6bd512fa9aa", "state_id": "grid_navigation:83485c159e2dc6bd512fa9aa", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . #\n. A # .\n# # # #\n. G . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.5, "west": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 0), (1, 1), (1, 2), (1, 3), (2, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 2], [2, 0], [2, 1], [2, 2], [2, 3], [3, 3]], "position": [1, 1], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.37, "west": 0.63}, "solvable": {"false": 0.71, "true": 0.29}, "value": {"0": 0.07, "1": 0.31, "2": 0.21, "3": 0.41}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:9bef3f9e09b636b7d3c57784", "state_id": "grid_navigation:9bef3f9e09b636b7d3c57784", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . G\n. . . .\n. . A .\n. # # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 0.0, "east": 0.5, "north": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 3), (1, 3), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[3, 1], [3, 2], [3, 3]], "position": [2, 2], "goal": [0, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.03, "north": 0.91, "west": 0.06}, "solvable": {"false": 0.06000000000000005, "true": 0.94}, "value": {"0": 0.05, "1": 0.82, "2": 0.12, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:8f1c21cae865bfd5d20ada6a", "state_id": "grid_navigation:8f1c21cae865bfd5d20ada6a", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A . #\n. . . .\n. . G .\n# # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.5, "west": 0.0, "east": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (0, 2), (0, 3), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [3, 0], [3, 1], [3, 2]], "position": [0, 1], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.02, "south": 0.89, "west": 0.09}, "solvable": {"false": 0.06999999999999995, "true": 0.93}, "value": {"0": 0.04, "1": 0.88, "2": 0.07, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:1de1d6b093b5db7d9f9c74bd", "state_id": "grid_navigation:1de1d6b093b5db7d9f9c74bd", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n# . A .\n# . . #\nG # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.3333333333333333, "west": 0.3333333333333333, "south": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 2), (1, 0), (1, 3), (2, 0), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 0], [2, 0], [2, 3], [3, 1], [3, 2]], "position": [1, 2], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.09, "south": 0.53, "west": 0.38}, "solvable": {"false": 0.35, "true": 0.65}, "value": {"0": 0.02, "1": 0.51, "2": 0.28, "3": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:b044166d4a25004955dee291", "state_id": "tic_tac_toe:b044166d4a25004955dee291", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X .\nX X O\nO . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_8": "Place O in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_8", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_9": 0.0, "cell_1": 0.0, "cell_3": 0.0, "cell_8": 1.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OOXX.X.:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".X.XXOO..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.35, "cell_9": 0.15, "cell_3": 0.22, "cell_1": 0.28}, "solvable": {"false": 0.85, "true": 0.15}, "value": {"0": 0.6, "1": 0.22, "2": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:67ef8c339bac5def2d10ff3a", "state_id": "grid_navigation:67ef8c339bac5def2d10ff3a", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # #\n. # # .\n. # . A\n# G . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 1.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 3), (1, 1), (2, 1), (2, 2), (3, 1), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [0, 3], [1, 1], [1, 2], [2, 1], [3, 0], [3, 3]], "position": [2, 3], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.27, "west": 0.73}, "solvable": {"false": 0.32999999999999996, "true": 0.67}, "value": {"0": 0.15, "1": 0.55, "2": 0.18, "3": 0.12}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:6b9c80946fc3e587128d02ed", "state_id": "grid_navigation:6b9c80946fc3e587128d02ed", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# G . A\n. . . .\n. . . #\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [2, 3]], "position": [0, 3], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.05, "west": 0.95}, "solvable": {"false": 0.19999999999999996, "true": 0.8}, "value": {"0": 0.56, "1": 0.38, "2": 0.04, "3": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:c55c1f47410d0da95225437b", "state_id": "grid_navigation:c55c1f47410d0da95225437b", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. # . #\nA . # .\n. G . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "south": 0.5, "east": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 2), (2, 1), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 1], [1, 3], [2, 2]], "position": [2, 0], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.11, "north": 0.08, "south": 0.81}, "solvable": {"false": 0.27, "true": 0.73}, "value": {"0": 0.56, "1": 0.29, "2": 0.1, "3": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:7c4315018c6f18ebbeda665d", "state_id": "tic_tac_toe:7c4315018c6f18ebbeda665d", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X O\nO . .\nX X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place O in row 2, column 3 (cell 6).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_6", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 1.0, "cell_5": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.OX.XXOX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XXOO..XXO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.25, "cell_5": 0.75}, "solvable": {"false": 0.84, "true": 0.16}, "value": {"0": 0.54, "1": 0.29, "2": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:49882b0fae9db4131ea9d147", "state_id": "grid_navigation:49882b0fae9db4131ea9d147", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\nA . . .\n. . G #\n# . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.0, "east": 0.5, "south": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (2, 3), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [2, 3], [3, 0]], "position": [1, 0], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.16, "north": 0.01, "east": 0.83}, "solvable": {"false": 0.06999999999999995, "true": 0.93}, "value": {"0": 0.11, "1": 0.81, "2": 0.07, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:2068775bb3da3381ab340ae0", "state_id": "grid_navigation:2068775bb3da3381ab340ae0", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\n# # . .\n. G # A\n. # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"south": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (1, 2), (2, 0), (2, 1), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 0], [1, 1], [2, 2], [3, 1]], "position": [2, 3], "goal": [2, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.43, "north": 0.57}, "solvable": {"false": 0.45999999999999996, "true": 0.54}, "value": {"0": 0.15, "1": 0.46, "2": 0.21, "3": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:3e10829094d20e75f815484a", "state_id": "grid_navigation:3e10829094d20e75f815484a", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . A #\n# # . #\nG # . #\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 2}, "gold_probs": {"action": {"west": 0.0, "south": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (0, 3), (1, 1), (1, 2), (3, 1), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 3], [1, 0], [1, 1], [1, 3], [2, 1], [2, 3]], "position": [0, 2], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.76, "west": 0.24}, "solvable": {"false": 0.32999999999999996, "true": 0.67}, "value": {"0": 0.02, "1": 0.47000000000000003, "2": 0.37, "3": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:e453793e49390ce3c820cb4f", "state_id": "tic_tac_toe:e453793e49390ce3c820cb4f", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\nX X .\nX O O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place O in row 1, column 2 (cell 2).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_6": "Place O in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_2": 0.25, "cell_3": 0.25, "cell_1": 0.25, "cell_6": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_2", "cell_3", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....XXOOX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "...XX.XOO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.4, "cell_1": 0.29, "cell_3": 0.24, "cell_2": 0.07}, "solvable": {"false": 0.77, "true": 0.23}, "value": {"0": 0.57, "1": 0.23, "2": 0.2}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:99cc55c7b64a912980ade36d", "state_id": "grid_navigation:99cc55c7b64a912980ade36d", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n. . A #\n. # . G\n# # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "south": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 0), (1, 3), (2, 0), (2, 1), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 3], [2, 1], [3, 0], [3, 1], [3, 2]], "position": [1, 2], "goal": [2, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.35, "south": 0.65}, "solvable": {"false": 0.31999999999999995, "true": 0.68}, "value": {"0": 0.17, "1": 0.58, "2": 0.12, "3": 0.13}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:c0b76cc99d5843fc9b05cc8b", "state_id": "tic_tac_toe:c0b76cc99d5843fc9b05cc8b", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X .\n. O .\nX X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_4": "Place O in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.25, "cell_3": 0.25, "cell_6": 0.25, "cell_4": 0.25}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_3", "cell_4", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXOX..X:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".X..O.XXO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.24, "cell_3": 0.17, "cell_6": 0.18, "cell_4": 0.41000000000000003}, "solvable": {"false": 0.8200000000000001, "true": 0.18}, "value": {"0": 0.47, "1": 0.31, "2": 0.22}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:d452f49e2ef788f992fa6e4b", "state_id": "tic_tac_toe:d452f49e2ef788f992fa6e4b", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O .\nO X X\n. . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place X in row 1, column 3 (cell 3).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_1": "Place X in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_3": 0.5, "cell_7": 0.0, "cell_8": 0.0, "cell_9": 0.5, "cell_1": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...OXX.O.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".O.OXX...", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.19, "cell_7": 0.18, "cell_1": 0.31, "cell_8": 0.18, "cell_9": 0.14}, "solvable": {"false": 0.69, "true": 0.31}, "value": {"0": 0.25, "1": 0.43, "2": 0.32}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:607bbc8294a4682aced44b03", "state_id": "tic_tac_toe:607bbc8294a4682aced44b03", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . X\nX . X\n. O O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place X in row 1, column 2 (cell 2).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.0, "cell_7": 0.0, "cell_5": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOX.XO.X:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.XX.X.OO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.37, "cell_2": 0.19, "cell_5": 0.44}, "solvable": {"false": 0.43999999999999995, "true": 0.56}, "value": {"0": 0.15, "1": 0.25, "2": 0.6}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:1df7e2c7def9521e9ae5db01", "state_id": "tic_tac_toe:1df7e2c7def9521e9ae5db01", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . X\nO O X\nX . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place O in row 3, column 2 (cell 8).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_1": "Place O in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_8": 0.0, "cell_9": 1.0, "cell_2": 0.0, "cell_1": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOOXX..:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..XOOXX..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.25, "cell_1": 0.18, "cell_8": 0.35, "cell_2": 0.22}, "solvable": {"false": 0.85, "true": 0.15}, "value": {"0": 0.55, "1": 0.26, "2": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:f3d4bc7bdf732d21b4b8e2d1", "state_id": "tic_tac_toe:f3d4bc7bdf732d21b4b8e2d1", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X .\nO O X\n. O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place O in row 1, column 3 (cell 3).", "cell_7": "Place O in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_3": 1.0, "cell_7": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXOOXXX.:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XX.OOX.OX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.48, "cell_7": 0.52}, "solvable": {"false": 0.84, "true": 0.16}, "value": {"0": 0.52, "1": 0.3, "2": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:9a9e30011193636f46d0e625", "state_id": "tic_tac_toe:9a9e30011193636f46d0e625", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . O\nX . .\nX . O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.5, "cell_6": 0.5, "cell_5": 0.0, "cell_8": 0.0, "cell_2": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OX..X.O:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..OX..X.O", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.45, "cell_6": 0.12, "cell_8": 0.2, "cell_1": 0.09, "cell_2": 0.14}, "solvable": {"false": 0.74, "true": 0.26}, "value": {"0": 0.21, "1": 0.38, "2": 0.41}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:9633b8aacf044cfebc070921", "state_id": "tic_tac_toe:9633b8aacf044cfebc070921", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O .\nO X X\n. X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place X in row 1, column 3 (cell 3).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_1": "Place X in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_3": 0.3333333333333333, "cell_7": 0.3333333333333333, "cell_1": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_3", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.O.OXX.XO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".O.OXX.XO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.38, "cell_1": 0.4, "cell_3": 0.22}, "solvable": {"false": 0.5700000000000001, "true": 0.43}, "value": {"0": 0.15, "1": 0.33, "2": 0.52}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:55e21473ee659389e35c0dc5", "state_id": "tic_tac_toe:55e21473ee659389e35c0dc5", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . .\nX . O\nX . X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place O in row 2, column 2 (cell 5).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_8", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_5": 0.0, "cell_8": 1.0, "cell_3": 0.0, "cell_2": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OO.XX.X:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O..X.OX.X", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.27, "cell_3": 0.12, "cell_5": 0.44, "cell_2": 0.17}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.44, "1": 0.34, "2": 0.22}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:8199d6688cba3b374c3d5844", "state_id": "grid_navigation:8199d6688cba3b374c3d5844", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\n. # . .\n. . . #\nA . G .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (2, 2), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 1], [2, 3]], "position": [3, 0], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.27, "east": 0.73}, "solvable": {"false": 0.09999999999999998, "true": 0.9}, "value": {"0": 0.47000000000000003, "1": 0.44, "2": 0.08, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:f64eccc9060c9010e8228ded", "state_id": "tic_tac_toe:f64eccc9060c9010e8228ded", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . X\n. O O\nO X X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place O in row 2, column 1 (cell 4).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 1.0, "cell_2": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.XXO.XOX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.X.OOOXX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.5, "cell_2": 0.5}, "solvable": {"false": 0.84, "true": 0.16}, "value": {"0": 0.36, "1": 0.45, "2": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:f9e5db9a01ab0cf4005da220", "state_id": "grid_navigation:f9e5db9a01ab0cf4005da220", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . A .\n# # # #\n# # . .\nG . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.5, "east": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 2), (0, 3), (1, 1), (1, 2), (2, 2), (3, 0), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 0], [1, 1], [1, 2], [1, 3], [2, 0], [2, 1], [3, 3]], "position": [0, 2], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.41, "east": 0.59}, "solvable": {"false": 0.43999999999999995, "true": 0.56}, "value": {"0": 0.01, "1": 0.36, "2": 0.38, "3": 0.25}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:62984e9a0323cf425225a961", "state_id": "tic_tac_toe:62984e9a0323cf425225a961", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . X\n. . X\nX O .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place O in row 2, column 1 (cell 4).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_4": 0.25, "cell_9": 0.25, "cell_5": 0.25, "cell_2": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_4", "cell_5", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXX..X.O:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.X..XXO.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.39, "cell_2": 0.09, "cell_4": 0.19, "cell_5": 0.33}, "solvable": {"false": 0.84, "true": 0.16}, "value": {"0": 0.57, "1": 0.24, "2": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:845e5569a6c18bceb2079a25", "state_id": "grid_navigation:845e5569a6c18bceb2079a25", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# G . .\n. . A .\n. # . .\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north.", "south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.0, "north": 0.5, "south": 0.0, "west": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (2, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [2, 1]], "position": [1, 2], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.01, "north": 0.47, "south": 0.01, "west": 0.51}, "solvable": {"false": 0.16000000000000003, "true": 0.84}, "value": {"0": 0.22, "1": 0.66, "2": 0.09, "3": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:0b6640fd799fb5430e92570f", "state_id": "grid_navigation:0b6640fd799fb5430e92570f", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . #\n. . # .\n. . # A\nG . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"south": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (1, 2), (2, 2), (3, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 3], [1, 2], [2, 2], [3, 2]], "position": [2, 3], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.31, "south": 0.69}, "solvable": {"false": 0.42000000000000004, "true": 0.58}, "value": {"0": 0.04, "1": 0.45, "2": 0.28, "3": 0.23}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:3fbd98205979a2769089c8d2", "state_id": "grid_navigation:3fbd98205979a2769089c8d2", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nA G # #\n. # # #\n. # . .\n. # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (0, 3), (1, 1), (1, 2), (1, 3), (2, 1), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [1, 1], [1, 2], [1, 3], [2, 1], [3, 1]], "position": [0, 0], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.45, "south": 0.55}, "solvable": {"false": 0.8, "true": 0.2}, "value": {"0": 0.35, "1": 0.06, "2": 0.08, "3": 0.51}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:62d13b3747ed7f6535246139", "state_id": "grid_navigation:62d13b3747ed7f6535246139", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\n. . . .\nG . A #\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "south": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 3), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [2, 3]], "position": [2, 2], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.06, "south": 0.02, "west": 0.92}, "solvable": {"false": 0.08999999999999997, "true": 0.91}, "value": {"0": 0.29, "1": 0.58, "2": 0.12, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:25e942fbf9347f8226d14a19", "state_id": "grid_navigation:25e942fbf9347f8226d14a19", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . .\nG . . .\n. A . .\n. . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "west": "Move one cell west.", "south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.0, "west": 0.5, "south": 0.0, "north": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [3, 3]], "position": [2, 1], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.01, "west": 0.52, "east": 0.01, "north": 0.46}, "solvable": {"false": 0.07999999999999996, "true": 0.92}, "value": {"0": 0.22, "1": 0.74, "2": 0.04, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:9e25a4d46d63a68d661bf3b9", "state_id": "tic_tac_toe:9e25a4d46d63a68d661bf3b9", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O O\n. . X\nX O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place O in row 2, column 2 (cell 5).", "cell_4": "Place O in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 1.0, "cell_4": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:OOXX..XOX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOO..XXOX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.5, "cell_4": 0.5}, "solvable": {"false": 0.8200000000000001, "true": 0.18}, "value": {"0": 0.37, "1": 0.46, "2": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:cc87d8e65df00b0c5a3acdb1", "state_id": "tic_tac_toe:cc87d8e65df00b0c5a3acdb1", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X O\n. . X\nO . X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place O in row 2, column 2 (cell 5).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_4": "Place O in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 0.3333333333333333, "cell_1": 0.3333333333333333, "cell_8": 0.0, "cell_4": 0.3333333333333333}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_4", "cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OX..OXX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XO..XO.X", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.2, "cell_1": 0.18, "cell_8": 0.3, "cell_5": 0.32}, "solvable": {"false": 0.88, "true": 0.12}, "value": {"0": 0.61, "1": 0.22, "2": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:59157bd015427ace0ccd3e76", "state_id": "grid_navigation:59157bd015427ace0ccd3e76", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\nG . # .\n# A . #\n# # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (0, 3), (1, 3), (2, 1), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 2], [2, 0], [2, 3], [3, 0], [3, 1]], "position": [2, 1], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.14, "north": 0.86}, "solvable": {"false": 0.4, "true": 0.6}, "value": {"0": 0.15, "1": 0.47, "2": 0.18, "3": 0.2}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:f8bd8b0142343d27294accc6", "state_id": "tic_tac_toe:f8bd8b0142343d27294accc6", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O .\nX X .\nO X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place O in row 2, column 3 (cell 6).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_6", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_6": 1.0, "cell_3": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OOXXXXO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XO.XX.OXO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.44, "cell_6": 0.56}, "solvable": {"false": 0.88, "true": 0.12}, "value": {"0": 0.46, "1": 0.37, "2": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:08a20bec7d9d4864e38cf92e", "state_id": "grid_navigation:08a20bec7d9d4864e38cf92e", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . A .\n. # . .\n# . G .\n# . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "south": 1.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 3), (1, 3), (2, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 1], [2, 0], [3, 0]], "position": [0, 2], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.18, "west": 0.19, "south": 0.63}, "solvable": {"false": 0.14, "true": 0.86}, "value": {"0": 0.03, "1": 0.74, "2": 0.19, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:3893544cb7a42e090bb42814", "state_id": "tic_tac_toe:3893544cb7a42e090bb42814", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . X\nO . O\n. . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_8": "Place X in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.0, "cell_2": 0.0, "cell_5": 1.0, "cell_1": 0.0, "cell_8": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XO.O..X:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..XO.O..X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.11, "cell_7": 0.15, "cell_1": 0.09, "cell_2": 0.13, "cell_5": 0.52}, "solvable": {"false": 0.64, "true": 0.36}, "value": {"0": 0.17, "1": 0.36, "2": 0.47}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:2b4992aac2011c6c98f25f59", "state_id": "tic_tac_toe:2b4992aac2011c6c98f25f59", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O .\n. X O\nX X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place O in row 1, column 3 (cell 3).", "cell_4": "Place O in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_3": 1.0, "cell_4": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOOXXX.X:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XO..XOXXO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.55, "cell_4": 0.45}, "solvable": {"false": 0.89, "true": 0.11}, "value": {"0": 0.6, "1": 0.29, "2": 0.11}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:207242c382f059b54a0e6030", "state_id": "grid_navigation:207242c382f059b54a0e6030", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\nA . # .\n# . # .\n# . . G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 3), (1, 1), (1, 3), (2, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 2], [2, 0], [2, 2], [3, 0]], "position": [1, 0], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.89, "north": 0.11}, "solvable": {"false": 0.21999999999999997, "true": 0.78}, "value": {"0": 0.01, "1": 0.63, "2": 0.28, "3": 0.08}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:d323c86646141c9199fd8266", "state_id": "grid_navigation:d323c86646141c9199fd8266", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n# . . .\n. # G #\nA . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 1), (1, 0), (1, 2), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [2, 1], [2, 3], [3, 2], [3, 3]], "position": [3, 0], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.5, "east": 0.5}, "solvable": {"false": 0.42000000000000004, "true": 0.58}, "value": {"0": 0.04, "1": 0.52, "2": 0.26, "3": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:ec4334f2565004b7e6f48c5c", "state_id": "tic_tac_toe:ec4334f2565004b7e6f48c5c", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O X\n. . .\nX O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_6": 0.0, "cell_9": 0.0, "cell_1": 0.0, "cell_5": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XO.OX..:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".OX...XO.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.28, "cell_6": 0.12, "cell_9": 0.27, "cell_4": 0.17, "cell_1": 0.16}, "solvable": {"false": 0.74, "true": 0.26}, "value": {"0": 0.2, "1": 0.45, "2": 0.35}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:f3d863bbf9a3c119665f628b", "state_id": "grid_navigation:f3d863bbf9a3c119665f628b", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # #\n# . A #\n. . . .\nG # . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 0.5, "south": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 3), (2, 0), (2, 3), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [1, 0], [1, 3], [3, 1], [3, 3]], "position": [1, 2], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.54, "south": 0.46}, "solvable": {"false": 0.18999999999999995, "true": 0.81}, "value": {"0": 0.01, "1": 0.61, "2": 0.3, "3": 0.08}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:4f885b7c26bad28c6edaca77", "state_id": "tic_tac_toe:4f885b7c26bad28c6edaca77", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X O\n. . O\n. . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_8": "Place X in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_9": 0.2, "cell_5": 0.2, "cell_7": 0.2, "cell_4": 0.2, "cell_8": 0.2}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4", "cell_5", "cell_7", "cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.....OXXO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XXO..O...", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.08, "cell_4": 0.25, "cell_7": 0.15, "cell_5": 0.39, "cell_9": 0.13}, "solvable": {"false": 0.84, "true": 0.16}, "value": {"0": 0.31, "1": 0.5, "2": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:19f210d0026ca2f1406fbcf6", "state_id": "tic_tac_toe:19f210d0026ca2f1406fbcf6", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X .\nX X .\nO . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place O in row 3, column 2 (cell 8).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_6": "Place O in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_8": 0.25, "cell_9": 0.25, "cell_3": 0.25, "cell_6": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_6", "cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....XXOXO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OX.XX.O..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.22, "cell_3": 0.28, "cell_6": 0.37, "cell_9": 0.13}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.67, "1": 0.15, "2": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:43a266f06b33c5063ac096e5", "state_id": "grid_navigation:43a266f06b33c5063ac096e5", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # .\n# . # G\n. . . A\n# . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "south": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (1, 1), (3, 0), (3, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [1, 0], [1, 2], [3, 0]], "position": [2, 3], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.7, "south": 0.05, "west": 0.25}, "solvable": {"false": 0.27, "true": 0.73}, "value": {"0": 0.11, "1": 0.6, "2": 0.18, "3": 0.11}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:06c90cc0b16bebeb60ce4731", "state_id": "tic_tac_toe:06c90cc0b16bebeb60ce4731", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O .\nO . .\nX . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place X in row 2, column 2 (cell 5).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_3": "Place X in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 0.2, "cell_8": 0.2, "cell_6": 0.2, "cell_1": 0.2, "cell_3": 0.2}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_3", "cell_5", "cell_6", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XO...OX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".O.O..X.X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.44, "cell_6": 0.12, "cell_1": 0.11, "cell_8": 0.24, "cell_3": 0.09}, "solvable": {"false": 0.63, "true": 0.37}, "value": {"0": 0.21, "1": 0.37, "2": 0.42}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:a03ea71bf9c75568d474842b", "state_id": "grid_navigation:a03ea71bf9c75568d474842b", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\nG # # A\n. . . .\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 1.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (1, 1), (2, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 1], [1, 2]], "position": [1, 3], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.12, "south": 0.88}, "solvable": {"false": 0.52, "true": 0.48}, "value": {"0": 0.04, "1": 0.34, "2": 0.48, "3": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:e12d4c79ccdc708bfeb75af9", "state_id": "tic_tac_toe:e12d4c79ccdc708bfeb75af9", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . O\nX O .\n. X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_7", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.0, "cell_7": 1.0, "cell_2": 0.0, "cell_9": 0.0, "cell_6": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXO..X.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..OXO..X.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.17, "cell_6": 0.27, "cell_1": 0.17, "cell_7": 0.14, "cell_9": 0.25}, "solvable": {"false": 0.74, "true": 0.26}, "value": {"0": 0.24, "1": 0.47000000000000003, "2": 0.29}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:b73167768b04018a75e3c6b2", "state_id": "tic_tac_toe:b73167768b04018a75e3c6b2", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\nO X .\n. X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_1": "Place X in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.3333333333333333, "cell_3": 0.3333333333333333, "cell_2": 0.3333333333333333, "cell_6": 0.0, "cell_1": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_3", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....XOOX.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "...OX..XO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.18, "cell_7": 0.27, "cell_2": 0.16, "cell_3": 0.2, "cell_1": 0.19}, "solvable": {"false": 0.69, "true": 0.31}, "value": {"0": 0.12, "1": 0.39, "2": 0.49}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:771a3cf603a77125f6118b62", "state_id": "tic_tac_toe:771a3cf603a77125f6118b62", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O X\n. . O\nO X X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_1": 0.3333333333333333, "cell_4": 0.3333333333333333, "cell_5": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_4", "cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OO.XXOX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".OX..OOXX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.31, "cell_1": 0.33, "cell_4": 0.36}, "solvable": {"false": 0.6, "true": 0.4}, "value": {"0": 0.27, "1": 0.34, "2": 0.39}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:db396cd23e7ba3d056ed4b6f", "state_id": "grid_navigation:db396cd23e7ba3d056ed4b6f", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG . . .\n. A # .\n. . . #\n. # . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "north": 0.5, "west": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((1, 2), (2, 3), (3, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 2], [2, 3], [3, 1], [3, 3]], "position": [1, 1], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.74, "west": 0.21, "south": 0.05}, "solvable": {"false": 0.29000000000000004, "true": 0.71}, "value": {"0": 0.55, "1": 0.31, "2": 0.09, "3": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:194890c32c405e0c6b8fb19a", "state_id": "tic_tac_toe:194890c32c405e0c6b8fb19a", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . X\n. X .\nO O X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place X in row 2, column 3 (cell 6).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_6", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 1.0, "cell_4": 0.0, "cell_2": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.O.XOX.X:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.X.X.OOX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.33, "cell_2": 0.26, "cell_4": 0.41}, "solvable": {"false": 0.53, "true": 0.47}, "value": {"0": 0.1, "1": 0.26, "2": 0.64}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:df3fc236a81f7dc0fb0f10fe", "state_id": "grid_navigation:df3fc236a81f7dc0fb0f10fe", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . #\n# . . A\nG . # .\n. . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 3), (2, 1), (3, 0), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 3], [1, 0], [2, 2], [3, 3]], "position": [1, 3], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.21, "west": 0.79}, "solvable": {"false": 0.24, "true": 0.76}, "value": {"0": 0.04, "1": 0.62, "2": 0.25, "3": 0.09}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:cfd40a22788b140b9d3b260c", "state_id": "tic_tac_toe:cfd40a22788b140b9d3b260c", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X .\n. O .\nX O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_1": "Place X in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.0, "cell_3": 0.0, "cell_6": 0.0, "cell_4": 0.0, "cell_1": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...OOXX..:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".X..O.XO.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.21, "cell_1": 0.16, "cell_3": 0.18, "cell_6": 0.16, "cell_9": 0.29}, "solvable": {"false": 0.62, "true": 0.38}, "value": {"0": 0.17, "1": 0.4, "2": 0.43}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:b282e2e13f95720954bb2526", "state_id": "grid_navigation:b282e2e13f95720954bb2526", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n# # . .\n. . . .\nA . . G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((1, 3), (2, 2), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 0], [1, 1]], "position": [3, 0], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.13, "east": 0.87}, "solvable": {"false": 0.09999999999999998, "true": 0.9}, "value": {"0": 0.05, "1": 0.75, "2": 0.17, "3": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:03fed40df9f704990515874e", "state_id": "grid_navigation:03fed40df9f704990515874e", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . .\n. . . .\n# # # .\n. . A G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((1, 1), (1, 2), (1, 3), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [2, 0], [2, 1], [2, 2]], "position": [3, 2], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.04, "east": 0.96}, "solvable": {"false": 0.050000000000000044, "true": 0.95}, "value": {"0": 0.99, "1": 0, "2": 0, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:746c96cd4201dfa54400c972", "state_id": "tic_tac_toe:746c96cd4201dfa54400c972", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X .\n. . .\nO X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place X in row 2, column 2 (cell 5).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_1": "Place X in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 0.3333333333333333, "cell_4": 0.0, "cell_6": 0.0, "cell_3": 0.3333333333333333, "cell_1": 0.3333333333333333}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_3", "cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OX.X..O:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".X....OXO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.12, "cell_4": 0.16, "cell_1": 0.13, "cell_3": 0.08, "cell_5": 0.51}, "solvable": {"false": 0.69, "true": 0.31}, "value": {"0": 0.17, "1": 0.46, "2": 0.37}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:68d9712747373b759f292566", "state_id": "grid_navigation:68d9712747373b759f292566", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. # . .\n. G # A\n# . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 1.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (1, 2), (2, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 1], [2, 2], [3, 0]], "position": [2, 3], "goal": [2, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.52, "north": 0.48}, "solvable": {"false": 0.43000000000000005, "true": 0.57}, "value": {"0": 0.17, "1": 0.54, "2": 0.17, "3": 0.12}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:f7a8df553666240cee711b1e", "state_id": "grid_navigation:f7a8df553666240cee711b1e", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\nA . . #\n. . . #\n# . G .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.5, "north": 0.0, "east": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 0), (2, 0), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 3], [2, 3], [3, 0]], "position": [1, 0], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.01, "east": 0.36, "south": 0.63}, "solvable": {"false": 0.10999999999999999, "true": 0.89}, "value": {"0": 0.01, "1": 0.7, "2": 0.23, "3": 0.06}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:8534004a62a0e0dba3a05dd3", "state_id": "tic_tac_toe:8534004a62a0e0dba3a05dd3", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O O\nX . .\nO X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place X in row 2, column 3 (cell 6).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.0, "cell_5": 1.0, "cell_9": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OX.OOXX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOOX..OX.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.34, "cell_5": 0.47, "cell_6": 0.19}, "solvable": {"false": 0.56, "true": 0.44}, "value": {"0": 0.19, "1": 0.36, "2": 0.45}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:6da1e96aef6302daa2788ccc", "state_id": "grid_navigation:6da1e96aef6302daa2788ccc", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n. . G .\n# . . A\n. . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.5, "south": 0.0, "west": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (2, 3), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [2, 0], [3, 2]], "position": [2, 3], "goal": [1, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.02, "west": 0.43, "north": 0.55}, "solvable": {"false": 0.15000000000000002, "true": 0.85}, "value": {"0": 0.26, "1": 0.63, "2": 0.08, "3": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:364611576bb3cd920fa71b0e", "state_id": "grid_navigation:364611576bb3cd920fa71b0e", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # G A\n. . # .\n. # . .\n# . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (0, 3), (1, 1), (2, 2), (3, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [1, 2], [2, 1], [3, 0], [3, 2]], "position": [0, 3], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.12, "west": 0.88}, "solvable": {"false": 0.56, "true": 0.44}, "value": {"0": 0.65, "1": 0.13, "2": 0.07, "3": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:4d028923d387e902d87c2c8d", "state_id": "tic_tac_toe:4d028923d387e902d87c2c8d", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . O\nX X .\nO X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.0, "cell_2": 0.5, "cell_6": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXX.OXO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.OXX.OX.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.14, "cell_6": 0.51, "cell_9": 0.35}, "solvable": {"false": 0.5, "true": 0.5}, "value": {"0": 0.2, "1": 0.28, "2": 0.52}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:3c76bc10f77be84f68760c8e", "state_id": "grid_navigation:3c76bc10f77be84f68760c8e", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # G .\n. . . #\n. # A .\n# # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.0, "south": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 0), (2, 2), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 3], [2, 1], [3, 0], [3, 1]], "position": [2, 2], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.71, "east": 0.22, "south": 0.07}, "solvable": {"false": 0.35, "true": 0.65}, "value": {"0": 0.04, "1": 0.56, "2": 0.28, "3": 0.12}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:1dbeed0eee468c9c4ce0ee81", "state_id": "tic_tac_toe:1dbeed0eee468c9c4ce0ee81", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X .\nO X .\nX O .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_1": 0.0, "cell_9": 0.0, "cell_6": 0.0, "cell_3": 1.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...OXXXO.:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".X.OX.XO.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.23, "cell_6": 0.25, "cell_3": 0.23, "cell_9": 0.29}, "solvable": {"false": 0.85, "true": 0.15}, "value": {"0": 0.62, "1": 0.21, "2": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:fc9dbd08bc4ba8e468b8e79a", "state_id": "grid_navigation:fc9dbd08bc4ba8e468b8e79a", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. G . .\n# . . #\n. . A .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 0.5, "north": 0.5, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 2), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[2, 0], [2, 3]], "position": [3, 2], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.02, "west": 0.19, "north": 0.79}, "solvable": {"false": 0.14, "true": 0.86}, "value": {"0": 0.05, "1": 0.71, "2": 0.18, "3": 0.06}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:e9943080d1649f2652017758", "state_id": "grid_navigation:e9943080d1649f2652017758", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG # # .\nA . . #\n. # . .\n# . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 1.0, "south": 0.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 2), (1, 3), (2, 1), (3, 0), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [1, 3], [2, 1], [3, 0], [3, 2]], "position": [1, 0], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.06, "north": 0.34, "east": 0.6}, "solvable": {"false": 0.43000000000000005, "true": 0.57}, "value": {"0": 0.1, "1": 0.35, "2": 0.23, "3": 0.32}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:b2595b1cec27534d74b11478", "state_id": "grid_navigation:b2595b1cec27534d74b11478", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A G #\n. # # #\n. . . #\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (1, 0), (1, 1), (1, 2), (2, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 1], [1, 2], [1, 3], [2, 3]], "position": [0, 1], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.02, "east": 0.98}, "solvable": {"false": 0.06999999999999995, "true": 0.93}, "value": {"0": 0.93, "1": 0.02, "2": 0.02, "3": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:df2bf525582ff4b25349f9b1", "state_id": "tic_tac_toe:df2bf525582ff4b25349f9b1", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X .\nO O X\nO X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.0, "cell_3": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.X.XOXOOX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XX.OOXOX.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.61, "cell_9": 0.39}, "solvable": {"false": 0.81, "true": 0.19}, "value": {"0": 0.3, "1": 0.44, "2": 0.26}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:558f5b0335d2e78af6ee1198", "state_id": "tic_tac_toe:558f5b0335d2e78af6ee1198", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X O\nX . .\nX . O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_6": "Place O in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.5, "cell_5": 0.0, "cell_8": 0.0, "cell_6": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.XOX..X.O:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XOX..X.O", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.25, "cell_5": 0.41, "cell_1": 0.18, "cell_6": 0.16}, "solvable": {"false": 0.89, "true": 0.11}, "value": {"0": 0.67, "1": 0.22, "2": 0.11}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:3cd632c1c5076cd8f255067f", "state_id": "grid_navigation:3cd632c1c5076cd8f255067f", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nA . # #\n. # # .\n# . . .\n. . . G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"south": 0.5, "east": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 3), (1, 2), (1, 3), (2, 2), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [1, 1], [1, 2], [2, 0]], "position": [0, 0], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.9, "east": 0.1}, "solvable": {"false": 0.25, "true": 0.75}, "value": {"0": 0.01, "1": 0.3, "2": 0.59, "3": 0.1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:8fe3a4537e357eb217f8ed5e", "state_id": "grid_navigation:8fe3a4537e357eb217f8ed5e", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n# # # .\n. # . .\n. # A G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 1.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (1, 2), (2, 1), (2, 2), (2, 3), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 0], [1, 1], [1, 2], [2, 1], [3, 1]], "position": [3, 2], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.04, "east": 0.96}, "solvable": {"false": 0.10999999999999999, "true": 0.89}, "value": {"0": 0.97, "1": 0.01, "2": 0.01, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:62f82e715017ed8577a5d43d", "state_id": "grid_navigation:62f82e715017ed8577a5d43d", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . A #\n. . . #\n# # # .\n. . . G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.5, "south": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (0, 3), (1, 1), (2, 1), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 3], [2, 0], [2, 1], [2, 2]], "position": [0, 2], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.78, "west": 0.22}, "solvable": {"false": 0.37, "true": 0.63}, "value": {"0": 0.01, "1": 0.33, "2": 0.51, "3": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:576efeaa04f572a1810ce71e", "state_id": "tic_tac_toe:576efeaa04f572a1810ce71e", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\n. X .\nO X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_4": "Place X in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.0, "cell_3": 0.0, "cell_6": 0.3333333333333333, "cell_2": 0.3333333333333333, "cell_4": 0.3333333333333333}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_4", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....X.OXO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "....X.OXO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.14, "cell_6": 0.17, "cell_4": 0.34, "cell_2": 0.2, "cell_3": 0.15}, "solvable": {"false": 0.78, "true": 0.22}, "value": {"0": 0.19, "1": 0.52, "2": 0.29}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:929a71292ebcc5342dd55e97", "state_id": "tic_tac_toe:929a71292ebcc5342dd55e97", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X X\nO O .\n. X X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_6": "Place O in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_6", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.5, "cell_6": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_6", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOXOXX.X:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXXOO..XX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.69, "cell_7": 0.31}, "solvable": {"false": 0.72, "true": 0.28}, "value": {"0": 0.32, "1": 0.31, "2": 0.37}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:a43dc9bd3910eaabcc37476b", "state_id": "tic_tac_toe:a43dc9bd3910eaabcc37476b", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X .\n. X O\n. . X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place O in row 3, column 2 (cell 8).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_4": "Place O in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_8", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_8": 1.0, "cell_3": 0.0, "cell_7": 0.0, "cell_4": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O.XXXO.:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OX..XO..X", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.32, "cell_7": 0.23, "cell_8": 0.22, "cell_4": 0.23}, "solvable": {"false": 0.9299999999999999, "true": 0.07}, "value": {"0": 0.82, "1": 0.12, "2": 0.06}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:de89b36e02377a42ba3b74cd", "state_id": "grid_navigation:de89b36e02377a42ba3b74cd", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # # .\n. # # .\n. . . A\nG . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.0, "west": 0.5, "south": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 3), (1, 2), (1, 3), (2, 2), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [0, 2], [1, 1], [1, 2]], "position": [2, 3], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.47, "south": 0.49, "north": 0.04}, "solvable": {"false": 0.15000000000000002, "true": 0.85}, "value": {"0": 0.05, "1": 0.76, "2": 0.15, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:9ce1dd9f498b3a273e86f74e", "state_id": "tic_tac_toe:9ce1dd9f498b3a273e86f74e", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . X\nX O .\nX . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place O in row 3, column 2 (cell 8).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_8": 0.3333333333333333, "cell_2": 0.3333333333333333, "cell_6": 0.0, "cell_9": 0.3333333333333333}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..X.O.XXO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.XXO.X..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.26, "cell_8": 0.39, "cell_2": 0.14, "cell_6": 0.21}, "solvable": {"false": 0.86, "true": 0.14}, "value": {"0": 0.73, "1": 0.13, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:493740c29e37d550b34c4133", "state_id": "grid_navigation:493740c29e37d550b34c4133", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\n. . A .\n. . . #\n. # G .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north.", "south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "north": 0.0, "south": 1.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 0), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [2, 3], [3, 1]], "position": [1, 2], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.03, "south": 0.69, "east": 0.06, "west": 0.22}, "solvable": {"false": 0.16000000000000003, "true": 0.84}, "value": {"0": 0.04, "1": 0.67, "2": 0.21, "3": 0.08}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:0270ff11134bf788ac8a495e", "state_id": "grid_navigation:0270ff11134bf788ac8a495e", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . #\n# . G .\nA . . #\n. . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 2), (1, 3), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 0], [2, 3], [3, 2]], "position": [2, 0], "goal": [1, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.03, "east": 0.97}, "solvable": {"false": 0.10999999999999999, "true": 0.89}, "value": {"0": 0.18, "1": 0.7, "2": 0.08, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:dca9c19c420104893c52b340", "state_id": "grid_navigation:dca9c19c420104893c52b340", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # A .\n. . G #\n. # . .\n. # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 1.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (1, 3), (2, 0), (2, 2), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 3], [2, 1], [3, 1], [3, 2]], "position": [0, 2], "goal": [1, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.55, "south": 0.45}, "solvable": {"false": 0.19999999999999996, "true": 0.8}, "value": {"0": 0.78, "1": 0.14, "2": 0.04, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:2d45b65ac43ab3b8d51433d2", "state_id": "tic_tac_toe:2d45b65ac43ab3b8d51433d2", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X .\n. X .\n. O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_4": 0.25, "cell_7": 0.25, "cell_3": 0.0, "cell_6": 0.25, "cell_9": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4", "cell_6", "cell_7", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...OXX..O:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OX..X..O.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.15, "cell_9": 0.21, "cell_4": 0.23, "cell_7": 0.19, "cell_3": 0.22}, "solvable": {"false": 0.64, "true": 0.36}, "value": {"0": 0.08, "1": 0.25, "2": 0.67}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:21d6c06ba6acb35a8d8e0ea4", "state_id": "grid_navigation:21d6c06ba6acb35a8d8e0ea4", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\n. G . A\n# # . .\n# . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "west": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (2, 0), (2, 1), (3, 0), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [2, 0], [2, 1], [3, 0], [3, 2], [3, 3]], "position": [1, 3], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.89, "north": 0.09, "south": 0.02}, "solvable": {"false": 0.24, "true": 0.76}, "value": {"0": 0.43, "1": 0.44, "2": 0.08, "3": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:dd8f88f3082a791e330c9977", "state_id": "tic_tac_toe:dd8f88f3082a791e330c9977", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O X\n. O O\n. X X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_7", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_1": 0.0, "cell_7": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...OOXXOX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".OX.OO.XX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.22, "cell_7": 0.33, "cell_1": 0.45}, "solvable": {"false": 0.47, "true": 0.53}, "value": {"0": 0.28, "1": 0.28, "2": 0.44}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:35e1676d6a7c085fa3f0b43a", "state_id": "tic_tac_toe:35e1676d6a7c085fa3f0b43a", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O X\n. X .\nO . X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place O in row 2, column 3 (cell 6).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_4": "Place O in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_6": 0.25, "cell_1": 0.25, "cell_8": 0.25, "cell_4": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_4", "cell_6", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OOX.X.X:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".OX.X.O.X", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.28, "cell_6": 0.18, "cell_8": 0.25, "cell_1": 0.29}, "solvable": {"false": 0.89, "true": 0.11}, "value": {"0": 0.59, "1": 0.27, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:fa172433c2860d63ad1d808e", "state_id": "tic_tac_toe:fa172433c2860d63ad1d808e", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . O\n. O .\nX X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place O in row 2, column 1 (cell 4).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_4": 0.25, "cell_6": 0.25, "cell_2": 0.25, "cell_9": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_4", "cell_6", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXO.X.X:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.O.O.XX.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.14, "cell_4": 0.16, "cell_2": 0.13, "cell_9": 0.57}, "solvable": {"false": 0.81, "true": 0.19}, "value": {"0": 0.47, "1": 0.25, "2": 0.28}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:3e76c017ca4a77a5a5a14996", "state_id": "tic_tac_toe:3e76c017ca4a77a5a5a14996", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . X\nO . .\nX . X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place O in row 3, column 2 (cell 8).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_8": 0.25, "cell_6": 0.25, "cell_5": 0.25, "cell_2": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_5", "cell_6", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.XO..X.X:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.XO..X.X", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.23, "cell_5": 0.44, "cell_6": 0.13, "cell_2": 0.2}, "solvable": {"false": 0.85, "true": 0.15}, "value": {"0": 0.46, "1": 0.31, "2": 0.23}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:5bd9d323a78ed9456ae3e093", "state_id": "grid_navigation:5bd9d323a78ed9456ae3e093", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\nA . . .\n. # G .\n. # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 1.0, "north": 0.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (0, 2), (1, 2), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [2, 1], [3, 1], [3, 2]], "position": [1, 0], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.89, "south": 0.08, "north": 0.03}, "solvable": {"false": 0.20999999999999996, "true": 0.79}, "value": {"0": 0.03, "1": 0.7, "2": 0.18, "3": 0.09}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:4e067cdfe8bfb833690dda9b", "state_id": "grid_navigation:4e067cdfe8bfb833690dda9b", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . #\n. . . G\n. . . #\nA . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.5, "east": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [2, 3]], "position": [3, 0], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.51, "north": 0.49}, "solvable": {"false": 0.08999999999999997, "true": 0.91}, "value": {"0": 0.02, "1": 0.67, "2": 0.25, "3": 0.06}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:741a5e47169ef0dddd506363", "state_id": "tic_tac_toe:741a5e47169ef0dddd506363", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O O\nX X O\n. X X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_7": "Place O in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 1.0, "cell_7": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOXXO.XX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".OOXXO.XX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.56, "cell_7": 0.44}, "solvable": {"false": 0.83, "true": 0.17}, "value": {"0": 0.52, "1": 0.28, "2": 0.2}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:b32a790912d3a56452a35e8c", "state_id": "tic_tac_toe:b32a790912d3a56452a35e8c", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . .\nX O .\n. X X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place O in row 1, column 2 (cell 2).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_6": "Place O in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_7", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_2": 0.0, "cell_7": 1.0, "cell_3": 0.0, "cell_6": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O.OXXX.:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O..XO..XX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.22, "cell_7": 0.3, "cell_6": 0.38, "cell_2": 0.1}, "solvable": {"false": 0.84, "true": 0.16}, "value": {"0": 0.51, "1": 0.25, "2": 0.24}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:4c01f4fc1787147847b19c46", "state_id": "tic_tac_toe:4c01f4fc1787147847b19c46", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X O\nX . O\n. O X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_1": "Place X in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.0, "cell_5": 0.0, "cell_1": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXX.O.XO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XOX.O.OX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.24, "cell_7": 0.26, "cell_5": 0.5}, "solvable": {"false": 0.77, "true": 0.23}, "value": {"0": 0.32, "1": 0.47, "2": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:af00c54a7dea915cfd55908e", "state_id": "grid_navigation:af00c54a7dea915cfd55908e", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. . G .\nA # . #\n. . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 2), (0, 3), (2, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[2, 1], [2, 3], [3, 3]], "position": [2, 0], "goal": [1, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.68, "south": 0.32}, "solvable": {"false": 0.28, "true": 0.72}, "value": {"0": 0.13, "1": 0.6, "2": 0.19, "3": 0.08}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:109bba4cfdaeaafba450d418", "state_id": "tic_tac_toe:109bba4cfdaeaafba450d418", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X .\n. . .\nO O X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_4": "Place X in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.25, "cell_6": 0.25, "cell_5": 0.0, "cell_3": 0.25, "cell_4": 0.25}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_3", "cell_4", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OX.O..X:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".X....OOX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.29, "cell_6": 0.27, "cell_1": 0.1, "cell_4": 0.22, "cell_3": 0.12}, "solvable": {"false": 0.5900000000000001, "true": 0.41}, "value": {"0": 0.22, "1": 0.38, "2": 0.4}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:ef837074d76b810a93af76f4", "state_id": "tic_tac_toe:ef837074d76b810a93af76f4", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . X\nX X .\n. . O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place O in row 3, column 2 (cell 8).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_8": 0.25, "cell_7": 0.25, "cell_6": 0.25, "cell_2": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_6", "cell_7", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXX.O.X:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.XXX...O", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.14, "cell_6": 0.51, "cell_2": 0.16, "cell_8": 0.19}, "solvable": {"false": 0.86, "true": 0.14}, "value": {"0": 0.52, "1": 0.29, "2": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:681a7f3f2af51e98eaf213c8", "state_id": "grid_navigation:681a7f3f2af51e98eaf213c8", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nA . . .\n. . . .\n. . . G\n# . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.5, "south": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[3, 0], [3, 2]], "position": [0, 0], "goal": [2, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.87, "south": 0.13}, "solvable": {"false": 0.06000000000000005, "true": 0.94}, "value": {"0": 0.01, "1": 0.76, "2": 0.21, "3": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:fac9f132f612efbe878c9a7e", "state_id": "tic_tac_toe:fac9f132f612efbe878c9a7e", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X .\nO X O\n. . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_7": 0.25, "cell_8": 0.25, "cell_3": 0.25, "cell_9": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_7", "cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...OXO.XX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XX.OXO...", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.45999999999999996, "cell_9": 0.15, "cell_8": 0.15, "cell_7": 0.24}, "solvable": {"false": 0.89, "true": 0.11}, "value": {"0": 0.49, "1": 0.35, "2": 0.16}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:b8d453aa106f8435e3840887", "state_id": "tic_tac_toe:b8d453aa106f8435e3840887", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . .\nO . X\nX O X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place X in row 1, column 2 (cell 2).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.0, "cell_3": 1.0, "cell_5": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OX.OXOX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O..O.XXOX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.5, "cell_2": 0.28, "cell_3": 0.22}, "solvable": {"false": 0.6699999999999999, "true": 0.33}, "value": {"0": 0.2, "1": 0.35, "2": 0.45}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:6108df53adab06b2fb62ca73", "state_id": "tic_tac_toe:6108df53adab06b2fb62ca73", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X X\n. X .\nO . O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place O in row 3, column 2 (cell 8).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_4": "Place O in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_8", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_8": 1.0, "cell_6": 0.0, "cell_1": 0.0, "cell_4": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXX.X.O:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XX.X.O.O", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.31, "cell_8": 0.24, "cell_6": 0.24, "cell_1": 0.21}, "solvable": {"false": 0.83, "true": 0.17}, "value": {"0": 0.43, "1": 0.32, "2": 0.25}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:f2ffd44a29fead2b297e7596", "state_id": "tic_tac_toe:f2ffd44a29fead2b297e7596", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . X\nO X X\n. . O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_8": "Place X in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_7", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 1.0, "cell_2": 0.0, "cell_8": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OOXXO.X:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.XOXX..O", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.58, "cell_2": 0.08, "cell_8": 0.34}, "solvable": {"false": 0.51, "true": 0.49}, "value": {"0": 0.14, "1": 0.29, "2": 0.57}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:efd108aaf3e7d1ccb1983104", "state_id": "grid_navigation:efd108aaf3e7d1ccb1983104", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . A\n. # # .\n. # . #\n. G # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 2}, "gold_probs": {"action": {"south": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (0, 3), (1, 1), (1, 3), (2, 1), (2, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 1], [1, 2], [2, 1], [2, 3], [3, 2], [3, 3]], "position": [0, 3], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.63, "south": 0.37}, "solvable": {"false": 0.52, "true": 0.48}, "value": {"0": 0.02, "1": 0.36, "2": 0.39, "3": 0.23}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:3995e7904ba0fa1e0a8131f4", "state_id": "tic_tac_toe:3995e7904ba0fa1e0a8131f4", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO O X\nX O X\n. . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place X in row 3, column 2 (cell 8).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_8": 0.0, "cell_9": 1.0, "cell_7": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...XOXOOX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OOXXOX...", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.15, "cell_9": 0.17, "cell_7": 0.68}, "solvable": {"false": 0.72, "true": 0.28}, "value": {"0": 0.17, "1": 0.64, "2": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:ba90dc3fe3bb0d01bd2b83d7", "state_id": "tic_tac_toe:ba90dc3fe3bb0d01bd2b83d7", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO O X\n. O X\nX . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_8": "Place X in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_9": 1.0, "cell_8": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XXO.XOO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OOX.OXX..", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.37, "cell_4": 0.33, "cell_9": 0.3}, "solvable": {"false": 0.61, "true": 0.39}, "value": {"0": 0.2, "1": 0.4, "2": 0.4}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:44f7c00489d46c489a04b2d4", "state_id": "tic_tac_toe:44f7c00489d46c489a04b2d4", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . O\nX O O\nX . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place X in row 1, column 2 (cell 2).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_1": "Place X in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.3333333333333333, "cell_8": 0.3333333333333333, "cell_1": 0.3333333333333333}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_2", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXOOX.X:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..OXOOX.X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.21, "cell_8": 0.57, "cell_1": 0.22}, "solvable": {"false": 0.55, "true": 0.45}, "value": {"0": 0.34, "1": 0.29, "2": 0.37}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:5301cbe194765303f48776b8", "state_id": "tic_tac_toe:5301cbe194765303f48776b8", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . O\n. X .\nX . X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place O in row 1, column 2 (cell 2).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_4": "Place O in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 1.0, "cell_8": 0.0, "cell_6": 0.0, "cell_4": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.O.X.X.X:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.O.X.X.X", "player": "O"}}}
{"id": "tic_tac_toe:01d564cfa968a8076c1e7639", "state_id": "tic_tac_toe:01d564cfa968a8076c1e7639", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . .\n. X .\nX O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place X in row 1, column 2 (cell 2).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.0, "cell_4": 0.3333333333333333, "cell_6": 0.3333333333333333, "cell_3": 0.3333333333333333, "cell_9": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_4", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....XOO.X:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O...X.XO.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.18, "cell_9": 0.34, "cell_6": 0.18, "cell_2": 0.13, "cell_4": 0.17}, "solvable": {"false": 0.61, "true": 0.39}, "value": {"0": 0.13, "1": 0.31, "2": 0.56}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:801da3aea9e5bbbab8d8f7df", "state_id": "tic_tac_toe:801da3aea9e5bbbab8d8f7df", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X .\n. . .\n. X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place X in row 1, column 3 (cell 3).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_3": 0.0, "cell_7": 0.0, "cell_6": 0.0, "cell_4": 0.0, "cell_5": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OX.XO..:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OX.....XO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.09, "cell_4": 0.09, "cell_5": 0.59, "cell_7": 0.12, "cell_3": 0.11}, "solvable": {"false": 0.71, "true": 0.29}, "value": {"0": 0.13, "1": 0.51, "2": 0.36}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:376fc76e86ae04e4581a9bd5", "state_id": "tic_tac_toe:376fc76e86ae04e4581a9bd5", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . O\nX O O\nX X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_1": "Place X in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.5, "cell_2": 0.0, "cell_1": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXOOXX.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..OXOOXX.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.71, "cell_2": 0.1, "cell_1": 0.19}, "solvable": {"false": 0.37, "true": 0.63}, "value": {"0": 0.16, "1": 0.17, "2": 0.67}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:daf7fddb0dbfae131b753f02", "state_id": "tic_tac_toe:daf7fddb0dbfae131b753f02", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O X\n. O .\n. . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place X in row 3, column 2 (cell 8).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_8", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_8": 1.0, "cell_9": 0.0, "cell_4": 0.0, "cell_7": 0.0, "cell_6": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....O.XOX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOX.O....", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.16, "cell_4": 0.36, "cell_9": 0.15, "cell_8": 0.19, "cell_7": 0.14}, "solvable": {"false": 0.84, "true": 0.16}, "value": {"0": 0.26, "1": 0.54, "2": 0.2}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:b0549fbc8efb4dcfbd194655", "state_id": "tic_tac_toe:b0549fbc8efb4dcfbd194655", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . X\n. X X\n. O O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.5, "cell_2": 0.0, "cell_7": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OOX.OXX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.X.XX.OO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.52, "cell_4": 0.38, "cell_2": 0.1}, "solvable": {"false": 0.29000000000000004, "true": 0.71}, "value": {"0": 0.09, "1": 0.16, "2": 0.75}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:bf5bc0652781b862cc7693ab", "state_id": "grid_navigation:bf5bc0652781b862cc7693ab", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n# . . .\n# . G #\n. A . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.5, "north": 0.5, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 1), (3, 1), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [2, 0], [2, 3], [3, 3]], "position": [3, 1], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.35, "north": 0.63, "west": 0.02}, "solvable": {"false": 0.13, "true": 0.87}, "value": {"0": 0.45, "1": 0.46, "2": 0.07, "3": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:f45bc60180c790b738cb4adc", "state_id": "grid_navigation:f45bc60180c790b738cb4adc", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nA . # #\n. . . .\n. . G .\n. # . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.5, "south": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 2), (3, 0), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [3, 1], [3, 3]], "position": [0, 0], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.86, "east": 0.14}, "solvable": {"false": 0.06000000000000005, "true": 0.94}, "value": {"0": 0.02, "1": 0.79, "2": 0.17, "3": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:8fc744f1163c9ab97866bb34", "state_id": "grid_navigation:8fc744f1163c9ab97866bb34", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG # . #\n. . . #\n. . A #\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 0.5, "south": 0.0, "north": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 3), (1, 3), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 3], [1, 3], [2, 3]], "position": [2, 2], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.37, "north": 0.51, "south": 0.12}, "solvable": {"false": 0.31000000000000005, "true": 0.69}, "value": {"0": 0.03, "1": 0.41, "2": 0.31, "3": 0.25}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:960e72abb0515dcb15d4a00f", "state_id": "grid_navigation:960e72abb0515dcb15d4a00f", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG . A .\n# # . #\n# . # .\n. . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "east": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 2), (1, 1), (2, 2), (2, 3), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [1, 1], [1, 3], [2, 0], [2, 2], [3, 2]], "position": [0, 2], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.17, "west": 0.65, "east": 0.18}, "solvable": {"false": 0.43000000000000005, "true": 0.57}, "value": {"0": 0.64, "1": 0.2, "2": 0.09, "3": 0.07}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:d502fd183193bd5885aae875", "state_id": "tic_tac_toe:d502fd183193bd5885aae875", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O O\n. . X\nO X X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place O in row 2, column 2 (cell 5).", "cell_4": "Place O in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 1.0, "cell_4": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.XX.OXXO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOO..XOXX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.6, "cell_5": 0.4}, "solvable": {"false": 0.81, "true": 0.19}, "value": {"0": 0.35, "1": 0.42, "2": 0.23}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:1e276febee5ddd56157492dd", "state_id": "grid_navigation:1e276febee5ddd56157492dd", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # . .\n. # G .\n. . # .\nA . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 2), (0, 3), (1, 2), (2, 1), (3, 0), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [1, 1], [2, 2], [3, 2], [3, 3]], "position": [3, 0], "goal": [1, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.43, "north": 0.57}, "solvable": {"false": 0.32999999999999996, "true": 0.67}, "value": {"0": 0.02, "1": 0.53, "2": 0.27, "3": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:e2af7494685ffe58d10af0d7", "state_id": "grid_navigation:e2af7494685ffe58d10af0d7", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\n# . . .\n. G # A\n# # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 1), (0, 2), (1, 2), (2, 0), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 0], [2, 2], [3, 0], [3, 1], [3, 2]], "position": [2, 3], "goal": [2, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.81, "south": 0.19}, "solvable": {"false": 0.26, "true": 0.74}, "value": {"0": 0.15, "1": 0.51, "2": 0.24, "3": 0.1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:9f58d237ca9d6c7c2d05983c", "state_id": "tic_tac_toe:9f58d237ca9d6c7c2d05983c", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O X\n. . .\nO . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place X in row 2, column 2 (cell 5).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_4": "Place X in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 0.3333333333333333, "cell_6": 0.3333333333333333, "cell_8": 0.0, "cell_1": 0.3333333333333333, "cell_4": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_5", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OO..X.X:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".OX...O.X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.18, "cell_8": 0.15, "cell_5": 0.27999999999999997, "cell_6": 0.23, "cell_1": 0.16}, "solvable": {"false": 0.7, "true": 0.3}, "value": {"0": 0.16, "1": 0.41, "2": 0.43}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:6f4cc5636af9386c41740a2d", "state_id": "grid_navigation:6f4cc5636af9386c41740a2d", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . G .\n. . . A\n# . . .\n. . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.5, "south": 0.0, "north": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((2, 3), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[2, 0], [3, 3]], "position": [1, 3], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.63, "south": 0.01, "west": 0.36}, "solvable": {"false": 0.08999999999999997, "true": 0.91}, "value": {"0": 0.15, "1": 0.8200000000000001, "2": 0.03, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:04171314e87fbef1e37218a2", "state_id": "tic_tac_toe:04171314e87fbef1e37218a2", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X .\nX . O\nO . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_9": 0.25, "cell_3": 0.25, "cell_1": 0.0, "cell_8": 0.25, "cell_5": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_5", "cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OO.X.X.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".X.X.OO..", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.11, "cell_8": 0.18, "cell_5": 0.46, "cell_9": 0.11, "cell_1": 0.14}, "solvable": {"false": 0.6799999999999999, "true": 0.32}, "value": {"0": 0.15, "1": 0.4, "2": 0.45}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:74d7f01fab893c798fba13f9", "state_id": "tic_tac_toe:74d7f01fab893c798fba13f9", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O X\n. X O\n. X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place O in row 2, column 1 (cell 4).", "cell_7": "Place O in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_7", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_4": 0.0, "cell_7": 1.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XXXOOOX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOX.XO.XO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.54, "cell_7": 0.46}, "solvable": {"false": 0.9, "true": 0.1}, "value": {"0": 0.49, "1": 0.41, "2": 0.1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:6871d13f469211de41f647c3", "state_id": "tic_tac_toe:6871d13f469211de41f647c3", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . O\nX X .\nO O X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place X in row 1, column 2 (cell 2).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.3333333333333333, "cell_1": 0.3333333333333333, "cell_6": 0.3333333333333333}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_2", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXX.OOX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..OXX.OOX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.56, "cell_2": 0.26, "cell_1": 0.18}, "solvable": {"false": 0.5, "true": 0.5}, "value": {"0": 0.16, "1": 0.34, "2": 0.5}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:45154004d29ce976f0d9112a", "state_id": "grid_navigation:45154004d29ce976f0d9112a", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . A\nG # # .\n# # # #\n. . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.5, "south": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 0), (1, 1), (1, 2), (2, 1), (2, 2), (2, 3), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 1], [1, 2], [2, 0], [2, 1], [2, 2], [2, 3], [3, 2], [3, 3]], "position": [0, 3], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.41, "west": 0.59}, "solvable": {"false": 0.71, "true": 0.29}, "value": {"0": 0.05, "1": 0.16, "2": 0.15, "3": 0.64}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:23f0de5ad72c307929bdb79f", "state_id": "grid_navigation:23f0de5ad72c307929bdb79f", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\nA . # .\n. . G #\n# . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.5, "south": 0.5, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 1), (0, 3), (1, 0), (2, 1), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 2], [2, 3], [3, 0], [3, 2], [3, 3]], "position": [1, 0], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.02, "east": 0.44, "south": 0.54}, "solvable": {"false": 0.12, "true": 0.88}, "value": {"0": 0.31, "1": 0.58, "2": 0.07, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:230a8ec5527e9496cabe71c0", "state_id": "grid_navigation:230a8ec5527e9496cabe71c0", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . A .\n. . . .\n. . # .\n. . G #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.5, "west": 0.5, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (1, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [2, 2], [3, 3]], "position": [0, 2], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.08, "west": 0.11, "south": 0.81}, "solvable": {"false": 0.10999999999999999, "true": 0.89}, "value": {"0": 0.01, "1": 0.56, "2": 0.36, "3": 0.07}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:a110525b11f5638c750cf4d1", "state_id": "tic_tac_toe:a110525b11f5638c750cf4d1", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . X\nO X .\nO O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place X in row 1, column 2 (cell 2).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.5, "cell_6": 0.0, "cell_9": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOX.OOX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.XOX.OO.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.49, "cell_6": 0.36, "cell_2": 0.15}, "solvable": {"false": 0.31000000000000005, "true": 0.69}, "value": {"0": 0.12, "1": 0.22, "2": 0.66}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:e07b316fddfba74a67c08f2f", "state_id": "tic_tac_toe:e07b316fddfba74a67c08f2f", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O O\nX O X\nX . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_1": "Place X in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.0, "cell_8": 0.0, "cell_1": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XXOXOO.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".OOXOXX..", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.5, "cell_9": 0.23, "cell_1": 0.27}, "solvable": {"false": 0.5700000000000001, "true": 0.43}, "value": {"0": 0.25, "1": 0.29, "2": 0.46}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:338d151b86633c9828fa76bf", "state_id": "tic_tac_toe:338d151b86633c9828fa76bf", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O .\nX . X\nX . O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place O in row 3, column 2 (cell 8).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_8": 0.25, "cell_3": 0.25, "cell_1": 0.25, "cell_5": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_3", "cell_5", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.O.X.XO.X:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".O.X.XX.O", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.25, "cell_5": 0.54, "cell_3": 0.07, "cell_1": 0.14}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.51, "1": 0.31, "2": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:2ee71d7216a0ede639c24523", "state_id": "tic_tac_toe:2ee71d7216a0ede639c24523", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . O\nX O .\nX X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 1.0, "cell_6": 0.0, "cell_2": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXO.XXO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.OXO.XX.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.4, "cell_2": 0.06, "cell_9": 0.54}, "solvable": {"false": 0.33999999999999997, "true": 0.66}, "value": {"0": 0.12, "1": 0.17, "2": 0.71}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:d8994006ed6621f8bd04fcc3", "state_id": "tic_tac_toe:d8994006ed6621f8bd04fcc3", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O O\nX X .\n. X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place O in row 2, column 3 (cell 6).", "cell_7": "Place O in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_6", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 1.0, "cell_7": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.XOXX.XOO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOOXX..XO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.65, "cell_7": 0.35}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.39, "1": 0.39999999999999997, "2": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:167fb1e438f4135e9b50e19f", "state_id": "tic_tac_toe:167fb1e438f4135e9b50e19f", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O .\nO . X\nX . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place X in row 1, column 3 (cell 3).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_3": 0.5, "cell_8": 0.0, "cell_1": 0.0, "cell_5": 0.0, "cell_9": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XX.O.O.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".O.O.XX..", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.13, "cell_8": 0.18, "cell_5": 0.48, "cell_9": 0.11, "cell_3": 0.1}, "solvable": {"false": 0.71, "true": 0.29}, "value": {"0": 0.23, "1": 0.44, "2": 0.33}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:ef4c79a6941fb57e11006ab6", "state_id": "tic_tac_toe:ef4c79a6941fb57e11006ab6", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . O\nO . .\nX X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place X in row 2, column 3 (cell 6).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.0, "cell_1": 0.0, "cell_2": 0.3333333333333333, "cell_5": 0.3333333333333333, "cell_9": 0.3333333333333333}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_5", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OO..XX.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..OO..XX.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.19, "cell_1": 0.06, "cell_9": 0.56, "cell_2": 0.08, "cell_6": 0.11}, "solvable": {"false": 0.43000000000000005, "true": 0.57}, "value": {"0": 0.1, "1": 0.17, "2": 0.73}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:40daca0f0df3905d4a0b1836", "state_id": "tic_tac_toe:40daca0f0df3905d4a0b1836", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . .\nO O .\nX O X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place X in row 1, column 3 (cell 3).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_3": 0.3333333333333333, "cell_6": 0.3333333333333333, "cell_2": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_3", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..X.OOXOX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X..OO.XOX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.51, "cell_2": 0.18, "cell_3": 0.31}, "solvable": {"false": 0.55, "true": 0.45}, "value": {"0": 0.2, "1": 0.32, "2": 0.48}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:4f4938da552551bb81ac84f3", "state_id": "grid_navigation:4f4938da552551bb81ac84f3", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # G #\n# . # .\n. A . .\n# # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.3333333333333333, "west": 0.3333333333333333, "east": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 2), (1, 1), (1, 3), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 3], [1, 0], [1, 2], [3, 0], [3, 1]], "position": [2, 1], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.71, "east": 0.18, "west": 0.11}, "solvable": {"false": 0.5900000000000001, "true": 0.41}, "value": {"0": 0.07, "1": 0.32, "2": 0.23, "3": 0.38}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:ca84a896ea86519c32217f04", "state_id": "tic_tac_toe:ca84a896ea86519c32217f04", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . .\nX . X\nO O .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.5, "cell_3": 0.0, "cell_5": 0.5, "cell_2": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XX.X.OO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X..X.XOO.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.24, "cell_3": 0.19, "cell_2": 0.21, "cell_5": 0.36}, "solvable": {"false": 0.74, "true": 0.26}, "value": {"0": 0.4, "1": 0.25, "2": 0.35}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:6a92abfa56072251b118f9ff", "state_id": "tic_tac_toe:6a92abfa56072251b118f9ff", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O .\nX . O\nX X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place O in row 1, column 3 (cell 3).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_3": 0.25, "cell_9": 0.25, "cell_1": 0.25, "cell_5": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_3", "cell_5", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.O.O.X.XX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".O.X.OXX.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.19, "cell_1": 0.07, "cell_3": 0.12, "cell_9": 0.62}, "solvable": {"false": 0.83, "true": 0.17}, "value": {"0": 0.57, "1": 0.2, "2": 0.23}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:d488df384a2576c932ed1136", "state_id": "grid_navigation:d488df384a2576c932ed1136", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. G . A\n# . . .\n. . . #\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [2, 3]], "position": [0, 3], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.89, "south": 0.11}, "solvable": {"false": 0.16000000000000003, "true": 0.84}, "value": {"0": 0.53, "1": 0.4, "2": 0.05, "3": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:0fea102aca266314d670e52b", "state_id": "tic_tac_toe:0fea102aca266314d670e52b", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . O\nO . X\nX X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.5, "cell_5": 0.0, "cell_2": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.XOX..XOO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.OO.XXX.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.14, "cell_5": 0.5, "cell_9": 0.36}, "solvable": {"false": 0.39, "true": 0.61}, "value": {"0": 0.18, "1": 0.18, "2": 0.64}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:7aa5108a3291c7512b3c34b3", "state_id": "tic_tac_toe:7aa5108a3291c7512b3c34b3", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . .\nO . .\nX O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place O in row 1, column 2 (cell 2).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.0, "cell_6": 0.0, "cell_5": 1.0, "cell_3": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..X..OXOX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X..O..XOX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.18, "cell_2": 0.13, "cell_5": 0.56, "cell_3": 0.13}, "solvable": {"false": 0.88, "true": 0.12}, "value": {"0": 0.39, "1": 0.37, "2": 0.24}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:0b79bbb1ac7e0fa7137d0d8b", "state_id": "tic_tac_toe:0b79bbb1ac7e0fa7137d0d8b", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . O\nO X .\n. X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.25, "cell_6": 0.0, "cell_7": 0.25, "cell_2": 0.25, "cell_9": 0.25}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_2", "cell_7", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OOX..X.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..OOX..X.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.14, "cell_6": 0.29000000000000004, "cell_2": 0.21, "cell_9": 0.2, "cell_7": 0.16}, "solvable": {"false": 0.75, "true": 0.25}, "value": {"0": 0.26, "1": 0.42, "2": 0.32}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:146466a532e483e48a54393b", "state_id": "grid_navigation:146466a532e483e48a54393b", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG . . #\n# # . .\n# . # #\n. A # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 2), (1, 1), (2, 2), (2, 3), (3, 0), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 0], [1, 1], [2, 0], [2, 2], [2, 3], [3, 2]], "position": [3, 1], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.33, "north": 0.67}, "solvable": {"false": 0.44999999999999996, "true": 0.55}, "value": {"0": 0.02, "1": 0.35, "2": 0.42, "3": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:ca396362bcd4e83a91383f77", "state_id": "grid_navigation:ca396362bcd4e83a91383f77", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\n# # . G\n. . . #\nA . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.5, "east": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (0, 3), (2, 0), (2, 1), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 0], [1, 1], [2, 3], [3, 3]], "position": [3, 0], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.43, "east": 0.57}, "solvable": {"false": 0.19999999999999996, "true": 0.8}, "value": {"0": 0.01, "1": 0.5, "2": 0.38, "3": 0.11}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:6fdeefd14cf4f906d66a9e7c", "state_id": "tic_tac_toe:6fdeefd14cf4f906d66a9e7c", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X X\n. X .\nO O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place O in row 2, column 1 (cell 4).", "cell_6": "Place O in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 1.0, "cell_6": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.OOXXX.X:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXX.X.OOX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.29, "cell_4": 0.71}, "solvable": {"false": 0.89, "true": 0.11}, "value": {"0": 0.59, "1": 0.29, "2": 0.12}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:228a1dac577dc481ad8926f3", "state_id": "tic_tac_toe:228a1dac577dc481ad8926f3", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . O\nX O .\n. X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place X in row 2, column 3 (cell 6).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_7", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.0, "cell_2": 0.0, "cell_7": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.XOXO.X.O:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.OXO..XO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.36, "cell_2": 0.29, "cell_7": 0.35}, "solvable": {"false": 0.76, "true": 0.24}, "value": {"0": 0.39, "1": 0.4, "2": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:d1bac4b73c49bcc3c2982cca", "state_id": "tic_tac_toe:d1bac4b73c49bcc3c2982cca", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O X\nO . .\nX X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_9": 0.25, "cell_1": 0.25, "cell_6": 0.25, "cell_5": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_5", "cell_6", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XX.OXO.:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".OXO..XX.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.46, "cell_6": 0.14, "cell_1": 0.08, "cell_5": 0.32}, "solvable": {"false": 0.81, "true": 0.19}, "value": {"0": 0.46, "1": 0.3, "2": 0.24}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:93483ef1f62c94e0916fc092", "state_id": "grid_navigation:93483ef1f62c94e0916fc092", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\nA G . .\n. . . .\n. . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "east": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((2, 3),)", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[3, 2]], "position": [1, 0], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.04, "east": 0.85, "north": 0.11}, "solvable": {"false": 0.020000000000000018, "true": 0.98}, "value": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:56065d1f0afe02e8e8368711", "state_id": "grid_navigation:56065d1f0afe02e8e8368711", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # #\n. A . G\n# . # #\n# . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east.", "north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "east": 1.0, "north": 0.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 2), (1, 0), (1, 2), (1, 3), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [2, 0], [2, 2], [2, 3], [3, 0], [3, 2]], "position": [1, 1], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.01, "south": 0.02, "north": 0.02, "east": 0.95}, "solvable": {"false": 0.15000000000000002, "true": 0.85}, "value": {"0": 0.85, "1": 0.09, "2": 0.02, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:0440ad6d60962ab358144152", "state_id": "grid_navigation:0440ad6d60962ab358144152", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n. # . .\nG . # .\n. # . A", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.5, "west": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (1, 2), (2, 1), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 1], [2, 2], [3, 1]], "position": [3, 3], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.42, "west": 0.58}, "solvable": {"false": 0.47, "true": 0.53}, "value": {"0": 0.01, "1": 0.27, "2": 0.49, "3": 0.23}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:9348b85238b22ce1606ddee8", "state_id": "tic_tac_toe:9348b85238b22ce1606ddee8", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO O .\nX X .\n. . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_8": "Place X in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.0, "cell_6": 0.5, "cell_7": 0.0, "cell_3": 0.5, "cell_8": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....XO.XO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OO.XX....", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.25, "cell_8": 0.06, "cell_9": 0.06, "cell_6": 0.52, "cell_7": 0.11}, "solvable": {"false": 0.45999999999999996, "true": 0.54}, "value": {"0": 0.19, "1": 0.21, "2": 0.6}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:a56a049af6da540335a3eb7b", "state_id": "tic_tac_toe:a56a049af6da540335a3eb7b", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X O\nO . .\n. . X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.0, "cell_6": 0.0, "cell_8": 0.0, "cell_5": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XO..XXO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XXOO....X", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.12, "cell_5": 0.7, "cell_6": 0.09, "cell_7": 0.09}, "solvable": {"false": 0.85, "true": 0.15}, "value": {"0": 0.25, "1": 0.52, "2": 0.23}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:ac83afc6d6673736318475d3", "state_id": "tic_tac_toe:ac83afc6d6673736318475d3", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . O\n. . O\nX X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place O in row 1, column 2 (cell 2).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.0, "cell_5": 0.0, "cell_4": 0.0, "cell_9": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOX..X.X:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.O..OXX.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.21, "cell_4": 0.11, "cell_2": 0.1, "cell_9": 0.58}, "solvable": {"false": 0.8, "true": 0.2}, "value": {"0": 0.42, "1": 0.3, "2": 0.28}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:2f707f63811e7fbbd2f7c535", "state_id": "tic_tac_toe:2f707f63811e7fbbd2f7c535", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O O\nX X .\n. O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_6", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.3333333333333333, "cell_9": 0.3333333333333333, "cell_6": 0.3333333333333333}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_6", "cell_7", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OOXO.XX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOOXX..O.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.28, "cell_6": 0.5, "cell_7": 0.22}, "solvable": {"false": 0.52, "true": 0.48}, "value": {"0": 0.16, "1": 0.37, "2": 0.47}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:4ac0bea86292cbf220081bf1", "state_id": "tic_tac_toe:4ac0bea86292cbf220081bf1", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . X\nO X X\nO O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.0, "cell_2": 0.0, "cell_9": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOXXOO.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..XOXXOO.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.18, "cell_2": 0.15, "cell_9": 0.67}, "solvable": {"false": 0.33999999999999997, "true": 0.66}, "value": {"0": 0.12, "1": 0.22, "2": 0.66}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:f3c29dc3828337d8b9df8d3e", "state_id": "grid_navigation:f3c29dc3828337d8b9df8d3e", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # .\nG . . #\n. . # .\n# # . A", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.5, "west": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 3), (1, 3), (2, 0), (2, 2), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [1, 3], [2, 2], [3, 0], [3, 1]], "position": [3, 3], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.41, "north": 0.59}, "solvable": {"false": 0.30000000000000004, "true": 0.7}, "value": {"0": 0.01, "1": 0.29, "2": 0.58, "3": 0.12}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:37a7cab54fb50a9c11dded9a", "state_id": "tic_tac_toe:37a7cab54fb50a9c11dded9a", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X X\n. . X\nO . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_8": "Place O in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.5, "cell_5": 0.0, "cell_4": 0.5, "cell_8": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OX..XXO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXX..XO..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.2, "cell_9": 0.17, "cell_5": 0.44, "cell_8": 0.19}, "solvable": {"false": 0.88, "true": 0.12}, "value": {"0": 0.53, "1": 0.29, "2": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:46c0c34c08eb016236a39971", "state_id": "grid_navigation:46c0c34c08eb016236a39971", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\nG # . #\n. . # .\n# . A .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 1.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 1), (2, 0), (2, 2), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 1], [1, 3], [2, 2], [3, 0]], "position": [3, 2], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.87, "east": 0.13}, "solvable": {"false": 0.5700000000000001, "true": 0.43}, "value": {"0": 0.01, "1": 0.31, "2": 0.42, "3": 0.26}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:cd2fd001092e67c273ba2567", "state_id": "grid_navigation:cd2fd001092e67c273ba2567", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nA . # .\n. . . .\n# . G #\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.5, "east": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (1, 3), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [2, 0], [2, 3]], "position": [0, 0], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.17, "south": 0.83}, "solvable": {"false": 0.09999999999999998, "true": 0.9}, "value": {"0": 0.02, "1": 0.71, "2": 0.23, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:2f245386ae119d44d73867e9", "state_id": "tic_tac_toe:2f245386ae119d44d73867e9", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O .\nX O X\nO X X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.0, "cell_3": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.O.XOXOXX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".O.XOXOXX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.33, "cell_1": 0.67}, "solvable": {"false": 0.86, "true": 0.14}, "value": {"0": 0.43, "1": 0.36, "2": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:aca3a981eeb2b6ca26079408", "state_id": "grid_navigation:aca3a981eeb2b6ca26079408", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. . G A\n. # # .\n. # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "south": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((1, 2), (2, 2), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[2, 1], [2, 2], [3, 1]], "position": [1, 3], "goal": [1, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.04, "west": 0.95, "south": 0.01}, "solvable": {"false": 0.09999999999999998, "true": 0.9}, "value": {"0": 0.96, "1": 0.02, "2": 0.01, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:884d8ba36f9e5b777e3564c2", "state_id": "tic_tac_toe:884d8ba36f9e5b777e3564c2", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X X\nO O X\n. . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_1": 0.25, "cell_8": 0.25, "cell_7": 0.25, "cell_9": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_7", "cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...OOX.XX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XXOOX...", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.29, "cell_9": 0.24, "cell_8": 0.15, "cell_1": 0.32}, "solvable": {"false": 0.83, "true": 0.17}, "value": {"0": 0.4, "1": 0.33, "2": 0.27}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:f92a2bf5abd1d27f1b82234c", "state_id": "tic_tac_toe:f92a2bf5abd1d27f1b82234c", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . X\nO X X\nO O .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 1.0, "cell_2": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOXXOX.X:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.XOXXOO.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.3, "cell_9": 0.7}, "solvable": {"false": 0.6799999999999999, "true": 0.32}, "value": {"0": 0.5, "1": 0.18, "2": 0.32}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:bf79932f32c6790f714738f6", "state_id": "grid_navigation:bf79932f32c6790f714738f6", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . A\n# # . .\nG # # .\n. . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.5, "south": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (0, 3), (1, 1), (1, 2), (2, 0), (2, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 0], [1, 1], [2, 1], [2, 2], [3, 2]], "position": [0, 3], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.65, "west": 0.35}, "solvable": {"false": 0.45999999999999996, "true": 0.54}, "value": {"0": 0.01, "1": 0.34, "2": 0.44, "3": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:2d0d596cc87b0346badf7131", "state_id": "tic_tac_toe:2d0d596cc87b0346badf7131", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO O .\nO . X\n. X X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.5, "cell_3": 0.5, "cell_5": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOX.OXX.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OO.O.X.XX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.25, "cell_7": 0.25, "cell_5": 0.5}, "solvable": {"false": 0.52, "true": 0.48}, "value": {"0": 0.35, "1": 0.25, "2": 0.39999999999999997}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:48f920d65b3b0e1ccbeaa48e", "state_id": "grid_navigation:48f920d65b3b0e1ccbeaa48e", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\nA . . #\n. G # .\n. # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.5, "east": 0.5, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (1, 2), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 3], [2, 2], [3, 1]], "position": [1, 0], "goal": [2, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.03, "east": 0.33, "south": 0.64}, "solvable": {"false": 0.32999999999999996, "true": 0.67}, "value": {"0": 0.38, "1": 0.4, "2": 0.13, "3": 0.09}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:3e1fcd8b1c1a10857e23c23b", "state_id": "grid_navigation:3e1fcd8b1c1a10857e23c23b", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . G\n. # # .\nA . . .\n. . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.5, "south": 0.0, "east": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 3), (1, 1), (2, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 1], [1, 2], [3, 3]], "position": [2, 0], "goal": [0, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.47, "south": 0.02, "east": 0.51}, "solvable": {"false": 0.22999999999999998, "true": 0.77}, "value": {"0": 0.01, "1": 0.58, "2": 0.33, "3": 0.08}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:81baa05bbc548dc76349ea21", "state_id": "tic_tac_toe:81baa05bbc548dc76349ea21", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . .\n. X X\n. . O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place X in row 1, column 2 (cell 2).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_4": "Place X in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.2, "cell_3": 0.2, "cell_7": 0.2, "cell_8": 0.2, "cell_4": 0.2}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_3", "cell_4", "cell_7", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O.X.OX.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O...XX..O", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.17, "cell_3": 0.21, "cell_4": 0.29, "cell_2": 0.15, "cell_8": 0.18}, "solvable": {"false": 0.48, "true": 0.52}, "value": {"0": 0.15, "1": 0.26, "2": 0.59}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:63b67124e3a9f82836f4a01b", "state_id": "tic_tac_toe:63b67124e3a9f82836f4a01b", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O .\nX . X\n. . O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.3333333333333333, "cell_8": 0.0, "cell_5": 0.3333333333333333, "cell_3": 0.0, "cell_7": 0.3333333333333333}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_5", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OX.X.O.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".O.X.X..O", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.53, "cell_3": 0.1, "cell_7": 0.12, "cell_8": 0.15, "cell_1": 0.1}, "solvable": {"false": 0.71, "true": 0.29}, "value": {"0": 0.18, "1": 0.47, "2": 0.35}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:3b6f879fb855ba9a6e47c5fa", "state_id": "tic_tac_toe:3b6f879fb855ba9a6e47c5fa", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . X\n. O .\nO . X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place O in row 1, column 2 (cell 2).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_8": "Place O in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_2": 0.25, "cell_6": 0.25, "cell_4": 0.25, "cell_8": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_4", "cell_6", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.X.O.X.X:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.X.O.O.X", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.27, "cell_6": 0.13, "cell_8": 0.26, "cell_4": 0.34}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.39999999999999997, "1": 0.39, "2": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:a5a621bdcbb11e57ff48ed99", "state_id": "grid_navigation:a5a621bdcbb11e57ff48ed99", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n# # . .\nG . # A\n. # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.5, "south": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 0), (1, 2), (2, 1), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 0], [1, 1], [2, 2], [3, 1]], "position": [2, 3], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.44, "north": 0.56}, "solvable": {"false": 0.4, "true": 0.6}, "value": {"0": 0.12, "1": 0.45, "2": 0.3, "3": 0.13}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:b1aedafe15b6dfacfa95934e", "state_id": "tic_tac_toe:b1aedafe15b6dfacfa95934e", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . X\n. O .\nX . O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place X in row 1, column 2 (cell 2).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.0, "cell_8": 0.0, "cell_1": 1.0, "cell_4": 0.0, "cell_6": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..X.O.X.O:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..X.O.X.O", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.22, "cell_2": 0.14, "cell_4": 0.2, "cell_6": 0.19, "cell_8": 0.25}, "solvable": {"false": 0.76, "true": 0.24}, "value": {"0": 0.18, "1": 0.5, "2": 0.32}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:6f7b8cb95b9548d750df7911", "state_id": "grid_navigation:6f7b8cb95b9548d750df7911", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # # #\n. G A .\n. # # #\n# # . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 1), (0, 2), (0, 3), (2, 1), (2, 2), (2, 3), (3, 0), (3, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [0, 2], [0, 3], [2, 1], [2, 2], [2, 3], [3, 0], [3, 1], [3, 3]], "position": [1, 2], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.97, "east": 0.03}, "solvable": {"false": 0.10999999999999999, "true": 0.89}, "value": {"0": 0.95, "1": 0.02, "2": 0.01, "3": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:ec96645740caf565918c3a68", "state_id": "grid_navigation:ec96645740caf565918c3a68", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # A .\n. # . #\n# # # .\nG . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"south": 0.5, "east": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (1, 1), (1, 2), (1, 3), (2, 1), (3, 0), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 1], [1, 3], [2, 0], [2, 1], [2, 2], [3, 3]], "position": [0, 2], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.58, "south": 0.42}, "solvable": {"false": 0.43000000000000005, "true": 0.57}, "value": {"0": 0.01, "1": 0.32, "2": 0.49, "3": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:62334cc868baeb3a654e448b", "state_id": "tic_tac_toe:62334cc868baeb3a654e448b", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X .\n. O X\nX . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_4": "Place O in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 1.0, "cell_3": 0.0, "cell_8": 0.0, "cell_4": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XXO..XO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OX..OXX..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.24, "cell_3": 0.19, "cell_8": 0.33, "cell_9": 0.24}, "solvable": {"false": 0.88, "true": 0.12}, "value": {"0": 0.47, "1": 0.32, "2": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:29b18c2660ce0d4b3ca8b0cd", "state_id": "grid_navigation:29b18c2660ce0d4b3ca8b0cd", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n# . . A\n. . # #\n# # G #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 2), (0, 3), (1, 0), (1, 1), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [2, 2], [2, 3], [3, 0], [3, 1], [3, 3]], "position": [1, 3], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.15, "west": 0.85}, "solvable": {"false": 0.47, "true": 0.53}, "value": {"0": 0.02, "1": 0.41, "2": 0.29, "3": 0.28}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:f068240a18cd0f74bbbb3a7b", "state_id": "grid_navigation:f068240a18cd0f74bbbb3a7b", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # #\n. . . .\n# . . .\n. G A .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "west": 1.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [2, 0]], "position": [3, 2], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.01, "north": 0.04, "west": 0.95}, "solvable": {"false": 0.07999999999999996, "true": 0.92}, "value": {"0": 0.72, "1": 0.23, "2": 0.04, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:d21505ee171b1532d9e1b7d7", "state_id": "grid_navigation:d21505ee171b1532d9e1b7d7", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . #\n. . . #\nG A . .\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north.", "east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 1.0, "north": 0.0, "east": 0.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 3]], "position": [2, 1], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.99, "north": 0, "south": 0.01, "east": 0}, "solvable": {"false": 0.06000000000000005, "true": 0.94}, "value": {"0": 0.99, "1": 0.01, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:3c1d9b2e676d3b0f4bb29675", "state_id": "grid_navigation:3c1d9b2e676d3b0f4bb29675", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# G . #\n. . A .\n. # . #\n# . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west.", "south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.5, "west": 0.5, "south": 0.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 3), (2, 1), (2, 3), (3, 0), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 3], [2, 1], [2, 3], [3, 0], [3, 3]], "position": [1, 2], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.46, "east": 0.02, "west": 0.51, "south": 0.01}, "solvable": {"false": 0.25, "true": 0.75}, "value": {"0": 0.3, "1": 0.53, "2": 0.1, "3": 0.07}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:72baf1d915d4f667274a4235", "state_id": "tic_tac_toe:72baf1d915d4f667274a4235", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O .\nO . X\nX . O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place X in row 1, column 3 (cell 3).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_8": "Place X in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_3": 0.3333333333333333, "cell_5": 0.3333333333333333, "cell_8": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_5", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXX.OO.X:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XO.O.XX.O", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.63, "cell_8": 0.24, "cell_3": 0.13}, "solvable": {"false": 0.63, "true": 0.37}, "value": {"0": 0.15, "1": 0.48, "2": 0.37}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:9083f72c514757b2a4630e34", "state_id": "tic_tac_toe:9083f72c514757b2a4630e34", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O X\nX . O\n. X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_7", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_7": 1.0, "cell_5": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.XOX.OXOX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOXX.O.XO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.68, "cell_7": 0.32}, "solvable": {"false": 0.89, "true": 0.11}, "value": {"0": 0.33, "1": 0.51, "2": 0.16}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:4a0855876541fda95769d436", "state_id": "grid_navigation:4a0855876541fda95769d436", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. # . .\nA . . .\n. G # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "south": 0.5, "east": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (2, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 1], [3, 2]], "position": [2, 0], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.1, "east": 0.28, "south": 0.62}, "solvable": {"false": 0.13, "true": 0.87}, "value": {"0": 0.52, "1": 0.36, "2": 0.08, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:da66ab31de3e544ee19ab3eb", "state_id": "tic_tac_toe:da66ab31de3e544ee19ab3eb", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X X\n. O .\nO X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 1.0, "cell_6": 0.0, "cell_4": 0.0, "cell_9": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXOXX..:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XX.O.OX.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.27, "cell_9": 0.29, "cell_6": 0.15, "cell_4": 0.29}, "solvable": {"false": 0.85, "true": 0.15}, "value": {"0": 0.43, "1": 0.36, "2": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:e64e9487ee2e3a758dc51c59", "state_id": "grid_navigation:e64e9487ee2e3a758dc51c59", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # G\n. . . .\nA . . #\n# . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.5, "north": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (2, 0), (3, 0), (3, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [2, 3], [3, 0], [3, 2], [3, 3]], "position": [2, 0], "goal": [0, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.49, "north": 0.51}, "solvable": {"false": 0.10999999999999999, "true": 0.89}, "value": {"0": 0.01, "1": 0.65, "2": 0.29, "3": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:2b0c336e2e0ffd1de2e3159d", "state_id": "grid_navigation:2b0c336e2e0ffd1de2e3159d", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n. A . .\nG . . #\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north.", "south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.5, "north": 0.0, "south": 0.5, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((2, 3), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [2, 3]], "position": [1, 1], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.01, "north": 0.01, "south": 0.1, "west": 0.88}, "solvable": {"false": 0.07999999999999996, "true": 0.92}, "value": {"0": 0.46, "1": 0.49, "2": 0.04, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:391e17343635215f0283237a", "state_id": "grid_navigation:391e17343635215f0283237a", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . A\n# . # .\n. . . G\n. . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "south": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (1, 2), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [1, 2], [3, 2]], "position": [0, 3], "goal": [2, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.53, "west": 0.47}, "solvable": {"false": 0.26, "true": 0.74}, "value": {"0": 0.07, "1": 0.61, "2": 0.2, "3": 0.12}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:5c7948b61b44f7a6447686dd", "state_id": "tic_tac_toe:5c7948b61b44f7a6447686dd", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X O\nX X .\nO . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place X in row 3, column 2 (cell 8).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_6", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_8": 0.3333333333333333, "cell_6": 0.3333333333333333, "cell_9": 0.3333333333333333}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_6", "cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O.XXOXO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXOXX.O..", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.17, "cell_6": 0.65, "cell_9": 0.18}, "solvable": {"false": 0.4, "true": 0.6}, "value": {"0": 0.08, "1": 0.15, "2": 0.77}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:23e3e2665dd3fb472f4db33a", "state_id": "tic_tac_toe:23e3e2665dd3fb472f4db33a", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . O\nX X O\n. X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_7": "Place O in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.5, "cell_1": 0.0, "cell_2": 0.5, "cell_7": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXXO.X.:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..OXXO.X.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.31, "cell_1": 0.28, "cell_7": 0.27, "cell_2": 0.14}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.38, "1": 0.39, "2": 0.23}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:ae19bfc23e81175ad98a8a55", "state_id": "tic_tac_toe:ae19bfc23e81175ad98a8a55", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . .\nO O .\n. . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place X in row 2, column 3 (cell 6).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_6", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_6": 1.0, "cell_3": 0.0, "cell_2": 0.0, "cell_8": 0.0, "cell_7": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..X.O.XO.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X..OO...X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.35, "cell_3": 0.15, "cell_2": 0.21, "cell_7": 0.14, "cell_8": 0.15}, "solvable": {"false": 0.6599999999999999, "true": 0.34}, "value": {"0": 0.22, "1": 0.36, "2": 0.42}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:06df4d5b1c518a5a23d102aa", "state_id": "tic_tac_toe:06df4d5b1c518a5a23d102aa", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O X\nX . .\n. O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place X in row 2, column 2 (cell 5).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 1.0, "cell_6": 0.0, "cell_9": 0.0, "cell_1": 0.0, "cell_7": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XO.O.X.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".OXX...O.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.11, "cell_5": 0.49, "cell_1": 0.13, "cell_7": 0.13, "cell_9": 0.14}, "solvable": {"false": 0.77, "true": 0.23}, "value": {"0": 0.24, "1": 0.47, "2": 0.29}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:f97e86c3030e53fd32522e43", "state_id": "grid_navigation:f97e86c3030e53fd32522e43", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . A\n. . . .\n. # . .\n# . G #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.5, "west": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 3), (1, 2), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [2, 1], [3, 0], [3, 3]], "position": [0, 3], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.28, "south": 0.72}, "solvable": {"false": 0.18000000000000005, "true": 0.82}, "value": {"0": 0.01, "1": 0.51, "2": 0.41, "3": 0.07}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:35496c8993271df7940bee36", "state_id": "tic_tac_toe:35496c8993271df7940bee36", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X O\nX . X\nO X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place O in row 2, column 2 (cell 5).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 1.0, "cell_9": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.XOX.XOXO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXOX.XOX.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.73, "cell_5": 0.27}, "solvable": {"false": 0.85, "true": 0.15}, "value": {"0": 0.35, "1": 0.44, "2": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:6c3d98163d4ff66fc9c2c00f", "state_id": "tic_tac_toe:6c3d98163d4ff66fc9c2c00f", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . X\n. O X\nX O .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place O in row 2, column 1 (cell 4).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_9": 0.5, "cell_1": 0.0, "cell_2": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..X.OOXX.:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..X.OXXO.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.64, "cell_2": 0.07, "cell_4": 0.15, "cell_1": 0.14}, "solvable": {"false": 0.83, "true": 0.17}, "value": {"0": 0.5, "1": 0.32, "2": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:a2ae3329d7597b3c3a60279e", "state_id": "grid_navigation:a2ae3329d7597b3c3a60279e", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\n# . A G\n# # . #\n. . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east.", "north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "east": 1.0, "north": 0.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (0, 3), (1, 3), (2, 0), (2, 2), (3, 1), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 0], [2, 0], [2, 1], [2, 3], [3, 2], [3, 3]], "position": [1, 2], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.01, "south": 0, "north": 0.01, "east": 0.98}, "solvable": {"false": 0.040000000000000036, "true": 0.96}, "value": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:2a71fce857a24354a5bbc08f", "state_id": "grid_navigation:2a71fce857a24354a5bbc08f", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nA . # .\n. . # .\nG . . .\n. . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.0, "south": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((2, 2), (2, 3), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 2], [3, 3]], "position": [0, 0], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.92, "east": 0.08}, "solvable": {"false": 0.12, "true": 0.88}, "value": {"0": 0.04, "1": 0.67, "2": 0.24, "3": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:430a530427450b2d74f91f14", "state_id": "tic_tac_toe:430a530427450b2d74f91f14", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O X\nX . .\n. X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place O in row 2, column 2 (cell 5).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_6": "Place O in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_5": 0.5, "cell_1": 0.0, "cell_7": 0.5, "cell_6": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXX...XO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".OXX...XO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.11, "cell_7": 0.14, "cell_5": 0.53, "cell_1": 0.22}, "solvable": {"false": 0.91, "true": 0.09}, "value": {"0": 0.34, "1": 0.45, "2": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:b4f3782d468b9ba5bc9b4e7c", "state_id": "grid_navigation:b4f3782d468b9ba5bc9b4e7c", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. G # .\n# # . .\n. # . #\n. A . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.5, "east": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 0), (1, 1), (2, 1), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 0], [1, 1], [2, 1], [2, 3]], "position": [3, 1], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.34, "east": 0.66}, "solvable": {"false": 0.44999999999999996, "true": 0.55}, "value": {"0": 0.02, "1": 0.38, "2": 0.39, "3": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:b42df0a21528a44b9e881394", "state_id": "grid_navigation:b42df0a21528a44b9e881394", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # G .\n# . # .\n# A # #\n# . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.5, "south": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 1), (1, 3), (2, 0), (2, 1), (2, 3), (3, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 0], [1, 2], [2, 0], [2, 2], [2, 3], [3, 0], [3, 2]], "position": [2, 1], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.84, "south": 0.16}, "solvable": {"false": 0.6799999999999999, "true": 0.32}, "value": {"0": 0.08, "1": 0.34, "2": 0.2, "3": 0.38}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:2bf7895c050f42a4323b81d0", "state_id": "grid_navigation:2bf7895c050f42a4323b81d0", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n# # . A\n. # . #\nG . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.5, "west": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (1, 1), (1, 2), (2, 0), (2, 3), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 0], [1, 1], [2, 1], [2, 3], [3, 2]], "position": [1, 3], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.79, "north": 0.21}, "solvable": {"false": 0.45999999999999996, "true": 0.54}, "value": {"0": 0.02, "1": 0.39, "2": 0.39, "3": 0.2}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:4902bcec62ae32181a673f56", "state_id": "tic_tac_toe:4902bcec62ae32181a673f56", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . O\nO X X\n. X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.0, "cell_2": 1.0, "cell_7": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOXX..XO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.OOXX.X.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.17, "cell_9": 0.32, "cell_7": 0.51}, "solvable": {"false": 0.6599999999999999, "true": 0.34}, "value": {"0": 0.48, "1": 0.27, "2": 0.25}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:4d4de4964f80853627024202", "state_id": "tic_tac_toe:4d4de4964f80853627024202", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X .\nO O .\n. X X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_6": "Place O in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_6", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.5, "cell_1": 0.0, "cell_3": 0.0, "cell_6": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_6", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XXOX.O.:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".X.OO..XX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.16, "cell_7": 0.27, "cell_6": 0.31, "cell_3": 0.26}, "solvable": {"false": 0.73, "true": 0.27}, "value": {"0": 0.47, "1": 0.2, "2": 0.33}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:dd5569ea9193e50f291536bb", "state_id": "grid_navigation:dd5569ea9193e50f291536bb", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # . #\n. . . .\nG # # A\n. # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 0), (1, 1), (1, 3), (2, 0), (2, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [0, 3], [2, 1], [2, 2], [3, 1], [3, 2]], "position": [2, 3], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.79, "south": 0.21}, "solvable": {"false": 0.51, "true": 0.49}, "value": {"0": 0.05, "1": 0.42, "2": 0.36, "3": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:29c6e02235dc5dc7e20cfdf7", "state_id": "tic_tac_toe:29c6e02235dc5dc7e20cfdf7", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\nO O X\nX X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_1": 0.0, "cell_3": 0.0, "cell_9": 1.0, "cell_2": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...OOXXX.:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "...OOXXX.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.04, "cell_3": 0.21, "cell_9": 0.67, "cell_1": 0.08}, "solvable": {"false": 0.8, "true": 0.2}, "value": {"0": 0.54, "1": 0.24, "2": 0.22}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:ceb25d06eb2df384d16988db", "state_id": "grid_navigation:ceb25d06eb2df384d16988db", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # . #\n# . . .\nG # . #\n# # A .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.5, "east": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 2), (0, 3), (1, 0), (1, 1), (1, 3), (3, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [0, 3], [1, 0], [2, 1], [2, 3], [3, 0], [3, 1]], "position": [3, 2], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.41, "north": 0.59}, "solvable": {"false": 0.43999999999999995, "true": 0.56}, "value": {"0": 0.03, "1": 0.45, "2": 0.34, "3": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:1d319cd7f2d7eaa681b03dd3", "state_id": "grid_navigation:1d319cd7f2d7eaa681b03dd3", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # .\n. . A .\n. . G #\n# # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 1.0, "east": 0.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (1, 3), (2, 0), (3, 0), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [2, 3], [3, 0], [3, 1]], "position": [1, 2], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.01, "south": 0.94, "west": 0.05}, "solvable": {"false": 0.07999999999999996, "true": 0.92}, "value": {"0": 0.92, "1": 0.06, "2": 0.01, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:834f132c1755d252c803a86e", "state_id": "tic_tac_toe:834f132c1755d252c803a86e", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X .\nX X O\n. . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_8": "Place O in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_8", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_7": 0.0, "cell_3": 0.0, "cell_9": 0.0, "cell_8": 1.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...OXX.XO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OX.XXO...", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.13, "cell_3": 0.42, "cell_9": 0.15, "cell_7": 0.3}, "solvable": {"false": 0.9, "true": 0.1}, "value": {"0": 0.69, "1": 0.17, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:97654b0595680ca4e6d5a1d5", "state_id": "tic_tac_toe:97654b0595680ca4e6d5a1d5", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X .\nX O X\nX O .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place O in row 1, column 3 (cell 3).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_3": 0.0, "cell_9": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXXOX.XO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OX.XOXXO.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.53, "cell_3": 0.47}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.59, "1": 0.28, "2": 0.13}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:2951fcc3ba53b26cd8d07635", "state_id": "tic_tac_toe:2951fcc3ba53b26cd8d07635", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . X\nX . .\nO . O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_8", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_1": 0.0, "cell_8": 1.0, "cell_2": 0.0, "cell_5": 0.0, "cell_6": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XX..O.O:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..XX..O.O", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.08, "cell_2": 0.1, "cell_8": 0.29, "cell_6": 0.13, "cell_5": 0.4}, "solvable": {"false": 0.6, "true": 0.4}, "value": {"0": 0.24, "1": 0.38, "2": 0.38}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:908ffd5257d9a6653867a06c", "state_id": "grid_navigation:908ffd5257d9a6653867a06c", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . .\n. A . .\n. # . #\n# G . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 0.0, "north": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (1, 1), (1, 3), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [2, 1], [2, 3], [3, 0]], "position": [1, 1], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.77, "north": 0.08, "east": 0.15}, "solvable": {"false": 0.30000000000000004, "true": 0.7}, "value": {"0": 0.03, "1": 0.63, "2": 0.25, "3": 0.09}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:f4c05f492b4eef216c2226b0", "state_id": "grid_navigation:f4c05f492b4eef216c2226b0", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\nA . # .\n# . # .\n. # . G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (1, 1), (1, 3), (2, 1), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 2], [2, 0], [2, 2], [3, 1]], "position": [1, 0], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.8, "north": 0.2}, "solvable": {"false": 0.45999999999999996, "true": 0.54}, "value": {"0": 0.01, "1": 0.35, "2": 0.45, "3": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:d9bb01e62094acacf1e47195", "state_id": "grid_navigation:d9bb01e62094acacf1e47195", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # #\n. . . .\n. . . A\nG . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 1.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 3), (3, 1), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [0, 3], [3, 3]], "position": [2, 3], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.92, "north": 0.08}, "solvable": {"false": 0.09999999999999998, "true": 0.9}, "value": {"0": 0.02, "1": 0.7, "2": 0.24, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:df0ca3081079a2b1c679e1bc", "state_id": "tic_tac_toe:df0ca3081079a2b1c679e1bc", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X O\nX O .\n. X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place X in row 2, column 3 (cell 6).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_6", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_6": 0.3333333333333333, "cell_7": 0.3333333333333333, "cell_9": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_6", "cell_7", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXOX.XO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXOXO..X.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.34, "cell_6": 0.42, "cell_7": 0.24}, "solvable": {"false": 0.64, "true": 0.36}, "value": {"0": 0.17, "1": 0.42, "2": 0.41}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:b936ec33221261af8e1d09ad", "state_id": "tic_tac_toe:b936ec33221261af8e1d09ad", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X O\nO . .\nX X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place X in row 2, column 3 (cell 6).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_1": "Place X in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.0, "cell_5": 1.0, "cell_1": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXX.XO.O:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XOO..XXO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.22, "cell_6": 0.2, "cell_5": 0.58}, "solvable": {"false": 0.5800000000000001, "true": 0.42}, "value": {"0": 0.15, "1": 0.28, "2": 0.57}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:c8d595f8ee05c150c04cf2b9", "state_id": "grid_navigation:c8d595f8ee05c150c04cf2b9", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A . #\n. . . G\n. . . .\n# . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 0.0, "east": 0.5, "south": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 3), (1, 3), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [3, 0], [3, 2], [3, 3]], "position": [0, 1], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.66, "west": 0.05, "east": 0.29}, "solvable": {"false": 0.06999999999999995, "true": 0.93}, "value": {"0": 0.35, "1": 0.58, "2": 0.04, "3": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:3f0b6a0adc0c199aa54f26cb", "state_id": "grid_navigation:3f0b6a0adc0c199aa54f26cb", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG # . #\n# . . .\n. A # #\n# . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.3333333333333333, "south": 0.3333333333333333, "west": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 3), (1, 0), (2, 2), (2, 3), (3, 0), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 3], [1, 0], [2, 2], [2, 3], [3, 0], [3, 2], [3, 3]], "position": [2, 1], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.66, "south": 0.08, "west": 0.26}, "solvable": {"false": 0.47, "true": 0.53}, "value": {"0": 0.05, "1": 0.47, "2": 0.25, "3": 0.23}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:987945fd05a0d6d7ccae05f6", "state_id": "grid_navigation:987945fd05a0d6d7ccae05f6", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . #\nG . . #\n. # A .\n# . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 0), (1, 2), (2, 3), (3, 0), (3, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 3], [1, 3], [2, 1], [3, 0], [3, 2], [3, 3]], "position": [2, 2], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.82, "east": 0.18}, "solvable": {"false": 0.35, "true": 0.65}, "value": {"0": 0.03, "1": 0.54, "2": 0.25, "3": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:07c5db098bbb1e32a7b9f042", "state_id": "tic_tac_toe:07c5db098bbb1e32a7b9f042", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X X\n. . O\n. . X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.0, "cell_8": 0.0, "cell_4": 1.0, "cell_5": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O..XXOX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXX..O..X", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.48, "cell_4": 0.26, "cell_8": 0.13, "cell_7": 0.13}, "solvable": {"false": 0.88, "true": 0.12}, "value": {"0": 0.5, "1": 0.34, "2": 0.16}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:d0312402b328fab606132dc2", "state_id": "tic_tac_toe:d0312402b328fab606132dc2", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . O\nX O X\n. X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place X in row 1, column 2 (cell 2).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_2": 0.3333333333333333, "cell_7": 0.3333333333333333, "cell_9": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_7", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.X.XOXO.O:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.OXOX.X.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.36, "cell_2": 0.11, "cell_7": 0.53}, "solvable": {"false": 0.6599999999999999, "true": 0.34}, "value": {"0": 0.32, "1": 0.38, "2": 0.3}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:360231e8257fcea8b392f336", "state_id": "grid_navigation:360231e8257fcea8b392f336", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# A . .\n. G # #\n. . # #\n# # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.0, "south": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 3), (1, 3), (2, 1), (2, 2), (2, 3), (3, 1), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 2], [1, 3], [2, 2], [2, 3], [3, 0], [3, 1], [3, 2]], "position": [0, 1], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.59, "east": 0.41}, "solvable": {"false": 0.14, "true": 0.86}, "value": {"0": 0.86, "1": 0.09, "2": 0.02, "3": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:ef8d28fd31ed61c7610b8e11", "state_id": "grid_navigation:ef8d28fd31ed61c7610b8e11", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # . .\n. . . .\n. . A #\n# # # G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.5, "west": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 2), (0, 3), (1, 0), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [2, 3], [3, 0], [3, 1], [3, 2]], "position": [2, 2], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.62, "north": 0.38}, "solvable": {"false": 0.48, "true": 0.52}, "value": {"0": 0.09, "1": 0.38, "2": 0.26, "3": 0.27}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:11d0e90fb53133fe99710420", "state_id": "tic_tac_toe:11d0e90fb53133fe99710420", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X .\nX . O\n. X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.0, "cell_3": 0.5, "cell_7": 0.0, "cell_5": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOX.X.X.:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".X.X.O.XO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.49, "cell_3": 0.14, "cell_1": 0.18, "cell_7": 0.19}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.52, "1": 0.35, "2": 0.13}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:b2645c1f1a9b24c7a18285db", "state_id": "tic_tac_toe:b2645c1f1a9b24c7a18285db", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O X\nO O .\n. . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place X in row 2, column 3 (cell 6).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_6", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 1.0, "cell_8": 0.0, "cell_7": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOO.XOX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOXOO...X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.22, "cell_6": 0.56, "cell_7": 0.22}, "solvable": {"false": 0.6799999999999999, "true": 0.32}, "value": {"0": 0.31, "1": 0.38, "2": 0.31}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:55be8e94edcb0a128ac3beeb", "state_id": "grid_navigation:55be8e94edcb0a128ac3beeb", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A . .\n. # . G\n# . # .\n. . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 2), (2, 1), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 1], [2, 0], [2, 2], [3, 3]], "position": [0, 1], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.08, "east": 0.92}, "solvable": {"false": 0.22999999999999998, "true": 0.77}, "value": {"0": 0.09, "1": 0.63, "2": 0.21, "3": 0.07}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:6207ce908164b6ad529f4484", "state_id": "tic_tac_toe:6207ce908164b6ad529f4484", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . O\n. . X\nO . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place X in row 2, column 2 (cell 5).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 1.0, "cell_2": 0.0, "cell_4": 0.0, "cell_8": 0.0, "cell_9": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OX..O.X:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.O..XO..", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.12, "cell_5": 0.47, "cell_8": 0.13, "cell_2": 0.12, "cell_9": 0.16}, "solvable": {"false": 0.78, "true": 0.22}, "value": {"0": 0.2, "1": 0.45, "2": 0.35}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:d3ced880880775b03f520388", "state_id": "grid_navigation:d3ced880880775b03f520388", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. # . .\n. . A G\n# # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "west": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (2, 0), (2, 2), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 1], [3, 0], [3, 1], [3, 2]], "position": [2, 2], "goal": [2, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.98, "north": 0.02, "west": 0}, "solvable": {"false": 0.040000000000000036, "true": 0.96}, "value": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:8c85220ffa68085b9577abbc", "state_id": "grid_navigation:8c85220ffa68085b9577abbc", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A # .\n# . # .\n. . . .\n. # G .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 1.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (2, 1), (2, 3), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 0], [1, 2], [3, 1]], "position": [0, 1], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.15, "south": 0.85}, "solvable": {"false": 0.21999999999999997, "true": 0.78}, "value": {"0": 0.02, "1": 0.58, "2": 0.33, "3": 0.07}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:28d794169b37ce4ec2a924ae", "state_id": "grid_navigation:28d794169b37ce4ec2a924ae", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # #\n# # . A\n. # . .\n. G # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": false, "value": 3}, "gold_probs": {"action": {"south": 0.5, "west": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (0, 3), (1, 1), (2, 0), (2, 1), (3, 1), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [0, 3], [1, 0], [1, 1], [2, 1], [3, 2], [3, 3]], "position": [1, 3], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.61, "west": 0.39}, "solvable": {"false": 0.4, "true": 0.6}, "value": {"0": 0.02, "1": 0.52, "2": 0.27, "3": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:530cb1870bca2c45ba419852", "state_id": "grid_navigation:530cb1870bca2c45ba419852", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # .\n. # # .\n. G . #\nA . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.5, "north": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 2), (1, 3), (2, 1), (2, 2), (3, 0), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [1, 1], [1, 2], [2, 3], [3, 2]], "position": [3, 0], "goal": [2, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.54, "north": 0.46}, "solvable": {"false": 0.36, "true": 0.64}, "value": {"0": 0.24, "1": 0.44, "2": 0.18, "3": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:2659eb681fca2e13b7af1716", "state_id": "grid_navigation:2659eb681fca2e13b7af1716", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n. G . .\n. . . A\n# . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.5, "west": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 2), (3, 0), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [3, 0], [3, 3]], "position": [2, 3], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.46, "north": 0.54}, "solvable": {"false": 0.08999999999999997, "true": 0.91}, "value": {"0": 0.08, "1": 0.79, "2": 0.11, "3": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:3c506ba3cc3be38509269de9", "state_id": "tic_tac_toe:3c506ba3cc3be38509269de9", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . O\nX X .\nO . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place O in row 3, column 2 (cell 8).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_8": 0.25, "cell_2": 0.25, "cell_6": 0.25, "cell_9": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_6", "cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O.X.OXX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.OXX.O..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.45, "cell_2": 0.27, "cell_8": 0.15, "cell_9": 0.13}, "solvable": {"false": 0.86, "true": 0.14}, "value": {"0": 0.55, "1": 0.25, "2": 0.2}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:fe8dd861d5ad67433aea81ec", "state_id": "tic_tac_toe:fe8dd861d5ad67433aea81ec", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X .\n. O .\nO . X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place O in row 2, column 3 (cell 6).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_8": "Place O in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.0, "cell_3": 1.0, "cell_4": 0.0, "cell_8": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XXO.X.O:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XX..O.O.X", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.15, "cell_8": 0.28, "cell_3": 0.34, "cell_4": 0.23}, "solvable": {"false": 0.83, "true": 0.17}, "value": {"0": 0.29, "1": 0.44, "2": 0.27}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:056df92a81bbbd8af1d6359b", "state_id": "grid_navigation:056df92a81bbbd8af1d6359b", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nA . # .\n. # . #\n. # # .\nG . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 1.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (1, 1), (1, 2), (2, 1), (2, 3), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 1], [1, 3], [2, 1], [2, 2], [3, 2]], "position": [0, 0], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.24, "south": 0.76}, "solvable": {"false": 0.4, "true": 0.6}, "value": {"0": 0.03, "1": 0.37, "2": 0.35, "3": 0.25}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:09726da16dffacb9199a3855", "state_id": "grid_navigation:09726da16dffacb9199a3855", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n. . . .\n# A . .\n. G . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 1.0, "east": 0.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [2, 0]], "position": [2, 1], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.07, "south": 0.91, "east": 0.02}, "solvable": {"false": 0.09999999999999998, "true": 0.9}, "value": {"0": 0.72, "1": 0.24, "2": 0.03, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:e822a7f568c71fcd6d36c3ab", "state_id": "grid_navigation:e822a7f568c71fcd6d36c3ab", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nA . . .\n. G . .\n. . . .\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.5, "east": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=()", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [], "position": [0, 0], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.37, "south": 0.63}, "solvable": {"false": 0.050000000000000044, "true": 0.95}, "value": {"0": 0.87, "1": 0.13, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:d93bca6ea2893462d7d7a56f", "state_id": "grid_navigation:d93bca6ea2893462d7d7a56f", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # #\n. # A G\n. . . #\n. # . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 2), (0, 3), (1, 0), (2, 1), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [1, 1], [2, 3], [3, 1], [3, 3]], "position": [1, 2], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.98, "south": 0.02}, "solvable": {"false": 0.07999999999999996, "true": 0.92}, "value": {"0": 0.98, "1": 0.01, "2": 0, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:ee20645e634ea2ca70563bf8", "state_id": "tic_tac_toe:ee20645e634ea2ca70563bf8", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO O .\n. X X\nX X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place O in row 1, column 3 (cell 3).", "cell_4": "Place O in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_3": 1.0, "cell_4": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOXX.OXX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OO..XXXXO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.49, "cell_4": 0.51}, "solvable": {"false": 0.85, "true": 0.15}, "value": {"0": 0.47, "1": 0.36, "2": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:aeb03d2459463efb87c1508b", "state_id": "tic_tac_toe:aeb03d2459463efb87c1508b", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X X\n. . O\n. . O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place X in row 3, column 2 (cell 8).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_8": 0.2, "cell_5": 0.2, "cell_4": 0.2, "cell_1": 0.2, "cell_7": 0.2}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_4", "cell_5", "cell_7", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.....XOOX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XX..O..O", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.13, "cell_5": 0.34, "cell_7": 0.13, "cell_4": 0.17, "cell_1": 0.23}, "solvable": {"false": 0.6599999999999999, "true": 0.34}, "value": {"0": 0.35, "1": 0.36, "2": 0.29}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:bcfffafc41867fd68609cffe", "state_id": "tic_tac_toe:bcfffafc41867fd68609cffe", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . O\nX X O\n. . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_7": 0.0, "cell_8": 0.0, "cell_2": 1.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XXXOO.O:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.OXXO..X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.31, "cell_2": 0.18, "cell_7": 0.51}, "solvable": {"false": 0.54, "true": 0.46}, "value": {"0": 0.22, "1": 0.34, "2": 0.44}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:f16bd61fbcf9e9d6a020bc45", "state_id": "grid_navigation:f16bd61fbcf9e9d6a020bc45", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . #\n. . . .\n. A . .\n. . . G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north.", "east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.5, "north": 0.0, "east": 0.5, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 3), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 3]], "position": [2, 1], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.54, "south": 0.45, "west": 0, "north": 0.01}, "solvable": {"false": 0.040000000000000036, "true": 0.96}, "value": {"0": 0.28, "1": 0.69, "2": 0.03, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:2f245744d65df0439620a622", "state_id": "tic_tac_toe:2f245744d65df0439620a622", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O X\nO . .\nX X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place O in row 2, column 3 (cell 6).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_6": 0.0, "cell_5": 1.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.XX.OXOX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOXO..XXO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.71, "cell_6": 0.29}, "solvable": {"false": 0.88, "true": 0.12}, "value": {"0": 0.5, "1": 0.35, "2": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:8347a5ba5f1bb4da370cfadb", "state_id": "tic_tac_toe:8347a5ba5f1bb4da370cfadb", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X X\nO . .\nO X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place O in row 2, column 2 (cell 5).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 0.0, "cell_1": 1.0, "cell_6": 0.0, "cell_9": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XX.XOO.:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XXO..OX.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.1, "cell_5": 0.47, "cell_1": 0.19, "cell_9": 0.24}, "solvable": {"false": 0.85, "true": 0.15}, "value": {"0": 0.4, "1": 0.36, "2": 0.24}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:3d5da4c7e801477a1a126aa9", "state_id": "grid_navigation:3d5da4c7e801477a1a126aa9", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . .\n. G . A\n. . . .\n# . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "north": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 3), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [3, 0], [3, 2]], "position": [1, 3], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.04, "north": 0.02, "west": 0.94}, "solvable": {"false": 0.06999999999999995, "true": 0.93}, "value": {"0": 0.71, "1": 0.26, "2": 0.02, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:3ef8ac3c8a0154a6261b971f", "state_id": "tic_tac_toe:3ef8ac3c8a0154a6261b971f", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . X\nX . .\nX O O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place O in row 2, column 2 (cell 5).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_6": "Place O in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_5": 0.25, "cell_1": 0.25, "cell_2": 0.25, "cell_6": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_2", "cell_5", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XX..XOO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..XX..XOO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.13, "cell_5": 0.41000000000000003, "cell_1": 0.22, "cell_2": 0.24}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.68, "1": 0.17, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:73ba05de93019093cf4fbeac", "state_id": "tic_tac_toe:73ba05de93019093cf4fbeac", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O .\n. . .\nO . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place X in row 2, column 2 (cell 5).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_4": "Place X in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 0.3333333333333333, "cell_8": 0.0, "cell_3": 0.3333333333333333, "cell_6": 0.3333333333333333, "cell_4": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_5", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XO..X.O:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XO....O.X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.52, "cell_4": 0.09, "cell_6": 0.08, "cell_8": 0.17, "cell_3": 0.14}, "solvable": {"false": 0.83, "true": 0.17}, "value": {"0": 0.18, "1": 0.57, "2": 0.25}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:9b46569a1906ee902c0d7043", "state_id": "tic_tac_toe:9b46569a1906ee902c0d7043", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X .\nX O X\nO X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.5, "cell_3": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.X.XOXOXO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OX.XOXOX.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.66, "cell_9": 0.34}, "solvable": {"false": 0.9, "true": 0.1}, "value": {"0": 0.44, "1": 0.42, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:4239d00955be6a7e0bfeb93b", "state_id": "grid_navigation:4239d00955be6a7e0bfeb93b", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\nG . # #\n. A # .\n. # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.5, "north": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 3), (2, 1), (2, 2), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 2], [1, 3], [2, 2], [3, 1]], "position": [2, 1], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.53, "north": 0.47}, "solvable": {"false": 0.39, "true": 0.61}, "value": {"0": 0.11, "1": 0.51, "2": 0.2, "3": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:7bb18dde5264bf05954b935d", "state_id": "tic_tac_toe:7bb18dde5264bf05954b935d", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O X\nX . O\nX O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.0, "cell_5": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXX.OXOX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".OXX.OXOX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.35, "cell_5": 0.65}, "solvable": {"false": 0.91, "true": 0.09}, "value": {"0": 0.43, "1": 0.41, "2": 0.16}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:6d44321e4c02b839583dbfe4", "state_id": "tic_tac_toe:6d44321e4c02b839583dbfe4", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . X\nX O .\n. O .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place O in row 1, column 2 (cell 2).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_6": "Place O in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 1.0, "cell_9": 0.0, "cell_7": 0.0, "cell_6": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOO..XX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.XXO..O.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.21, "cell_7": 0.29, "cell_2": 0.2, "cell_6": 0.3}, "solvable": {"false": 0.84, "true": 0.16}, "value": {"0": 0.55, "1": 0.23, "2": 0.22}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:0f24411b723ab46ae7163bbe", "state_id": "grid_navigation:0f24411b723ab46ae7163bbe", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. G # .\n. . . #\n. # # .\nA . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 3), (2, 1), (2, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 3], [2, 1], [2, 2], [3, 3]], "position": [3, 0], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.35, "north": 0.65}, "solvable": {"false": 0.25, "true": 0.75}, "value": {"0": 0.02, "1": 0.53, "2": 0.29, "3": 0.16}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:cb1d9600654f7e07516c481c", "state_id": "grid_navigation:cb1d9600654f7e07516c481c", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # A .\n. # . .\n. # . .\n. G # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "south": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 1), (2, 1), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 1], [2, 1], [3, 2]], "position": [0, 2], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.71, "south": 0.29}, "solvable": {"false": 0.6599999999999999, "true": 0.34}, "value": {"0": 0.02, "1": 0.27, "2": 0.46, "3": 0.25}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:5988c47360ae1c68246aa198", "state_id": "grid_navigation:5988c47360ae1c68246aa198", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A . .\n# . G .\n. . # .\n. . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.5, "east": 0.5, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 3), (1, 2), (1, 3), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [2, 2], [3, 2], [3, 3]], "position": [0, 1], "goal": [1, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.82, "south": 0.17, "west": 0.01}, "solvable": {"false": 0.12, "true": 0.88}, "value": {"0": 0.81, "1": 0.15, "2": 0.02, "3": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:f580b79c2c8bfd3b382e5e45", "state_id": "grid_navigation:f580b79c2c8bfd3b382e5e45", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # .\n. # G .\n. . . A\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "west": 0.5, "north": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (0, 2), (1, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [1, 1]], "position": [2, 3], "goal": [1, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.43, "north": 0.52, "south": 0.05}, "solvable": {"false": 0.16000000000000003, "true": 0.84}, "value": {"0": 0.07, "1": 0.69, "2": 0.19, "3": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:4a6e9c536bc9bf268234458a", "state_id": "grid_navigation:4a6e9c536bc9bf268234458a", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . G\n# . . #\n. # # .\nA . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (1, 2), (2, 2), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [1, 3], [2, 1], [2, 2]], "position": [3, 0], "goal": [0, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.67, "north": 0.33}, "solvable": {"false": 0.24, "true": 0.76}, "value": {"0": 0.01, "1": 0.42, "2": 0.45, "3": 0.12}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:9f97e162959da7bbe32b7c38", "state_id": "grid_navigation:9f97e162959da7bbe32b7c38", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. G # #\n# . # #\n# # . #\n. . A .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.3333333333333333, "west": 0.3333333333333333, "north": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (0, 3), (1, 0), (1, 2), (1, 3), (2, 0), (2, 1), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [1, 0], [1, 2], [1, 3], [2, 0], [2, 1], [2, 3]], "position": [3, 2], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.66, "west": 0.26, "east": 0.08}, "solvable": {"false": 0.6, "true": 0.4}, "value": {"0": 0.02, "1": 0.3, "2": 0.36, "3": 0.32}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:d122298216c8f23a6b48eb64", "state_id": "tic_tac_toe:d122298216c8f23a6b48eb64", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O O\nX X O\n. X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_1": 0.3333333333333333, "cell_9": 0.3333333333333333, "cell_7": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_7", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOXXO.X.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".OOXXO.X.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.28, "cell_7": 0.42, "cell_1": 0.3}, "solvable": {"false": 0.51, "true": 0.49}, "value": {"0": 0.2, "1": 0.26, "2": 0.54}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:6f104e7822fb687eb12029f0", "state_id": "tic_tac_toe:6f104e7822fb687eb12029f0", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . O\nX . O\nX . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place X in row 2, column 2 (cell 5).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_8": "Place X in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_8", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 0.0, "cell_2": 0.0, "cell_8": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.OO.XX.X:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.OX.OX.X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.25, "cell_2": 0.3, "cell_5": 0.45}, "solvable": {"false": 0.5, "true": 0.5}, "value": {"0": 0.19, "1": 0.29, "2": 0.52}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:8907151179f21260576d4f92", "state_id": "tic_tac_toe:8907151179f21260576d4f92", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O X\n. O .\nO X X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place O in row 2, column 1 (cell 4).", "cell_6": "Place O in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_6", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_4": 0.0, "cell_6": 1.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.XXOOX.X:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOX.O.OXX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.27, "cell_4": 0.73}, "solvable": {"false": 0.8200000000000001, "true": 0.18}, "value": {"0": 0.29, "1": 0.45, "2": 0.26}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:971609f7a7947681b6c78042", "state_id": "tic_tac_toe:971609f7a7947681b6c78042", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O O\n. O X\nX . X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place O in row 2, column 1 (cell 4).", "cell_8": "Place O in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_8", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_8": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:OOXXO.X.X:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOO.OXX.X", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.48, "cell_4": 0.52}, "solvable": {"false": 0.81, "true": 0.19}, "value": {"0": 0.38, "1": 0.37, "2": 0.25}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:d175274216408165691448ee", "state_id": "grid_navigation:d175274216408165691448ee", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A . #\n. . . .\n. . . G\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 0.0, "south": 0.5, "east": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3),)", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3]], "position": [0, 1], "goal": [2, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.08, "south": 0.9, "west": 0.02}, "solvable": {"false": 0.050000000000000044, "true": 0.95}, "value": {"0": 0.02, "1": 0.79, "2": 0.15, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:d6a65ea03848b7f850b2c066", "state_id": "tic_tac_toe:d6a65ea03848b7f850b2c066", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X X\nX O .\n. . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_6": "Place O in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.0, "cell_8": 0.0, "cell_9": 1.0, "cell_6": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....OXXXO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXXXO....", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.1, "cell_6": 0.5, "cell_9": 0.24, "cell_7": 0.16}, "solvable": {"false": 0.9, "true": 0.1}, "value": {"0": 0.47, "1": 0.35, "2": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:3dfd4cafbfe7d7298630c34c", "state_id": "tic_tac_toe:3dfd4cafbfe7d7298630c34c", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X .\nO X X\nX . O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place O in row 3, column 2 (cell 8).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_8": 0.5, "cell_3": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.XOXX.OOX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OX.OXXX.O", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.65, "cell_3": 0.35}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.66, "1": 0.2, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:91a89e4f859fe52ae00b92af", "state_id": "tic_tac_toe:91a89e4f859fe52ae00b92af", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O .\nO . X\nX X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_9": 1.0, "cell_1": 0.0, "cell_5": 0.0, "cell_3": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.O.O.XXX.:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".O.O.XXX.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.31, "cell_3": 0.12, "cell_9": 0.48, "cell_1": 0.09}, "solvable": {"false": 0.74, "true": 0.26}, "value": {"0": 0.44, "1": 0.24, "2": 0.32}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:20ead5182eabd8875fc65021", "state_id": "grid_navigation:20ead5182eabd8875fc65021", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # # .\n. # # #\n. G A .\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 1.0, "south": 0.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 3), (1, 2), (1, 3), (2, 2), (2, 3), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [0, 2], [1, 1], [1, 2], [1, 3]], "position": [2, 2], "goal": [2, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.03, "east": 0.01, "west": 0.96}, "solvable": {"false": 0.07999999999999996, "true": 0.92}, "value": {"0": 0.89, "1": 0.08, "2": 0.01, "3": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:2c0b3e045f3aedab245d6b29", "state_id": "grid_navigation:2c0b3e045f3aedab245d6b29", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A . .\n# # . .\n# . . .\n. G # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 1.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 0), (2, 0), (2, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [1, 1], [2, 0], [3, 2]], "position": [0, 1], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.25, "east": 0.75}, "solvable": {"false": 0.19999999999999996, "true": 0.8}, "value": {"0": 0.01, "1": 0.53, "2": 0.38, "3": 0.08}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:709f7290ad9baf2b54b3d151", "state_id": "tic_tac_toe:709f7290ad9baf2b54b3d151", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO O .\n. . X\n. X X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place O in row 1, column 3 (cell 3).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_3": 1.0, "cell_4": 0.0, "cell_7": 0.0, "cell_5": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OX.OXX.:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OO...X.XX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.24, "cell_3": 0.24, "cell_4": 0.15, "cell_5": 0.37}, "solvable": {"false": 0.8200000000000001, "true": 0.18}, "value": {"0": 0.49, "1": 0.29, "2": 0.22}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:bee3772becccff25a29284d7", "state_id": "grid_navigation:bee3772becccff25a29284d7", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . A G\n. . . .\n# . . #\n# # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "west": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (1, 3), (2, 3), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[2, 0], [2, 3], [3, 0], [3, 1], [3, 2]], "position": [0, 2], "goal": [0, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.03, "west": 0.03, "east": 0.94}, "solvable": {"false": 0.030000000000000027, "true": 0.97}, "value": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:1e3f9571af4da38a8585f3a0", "state_id": "tic_tac_toe:1e3f9571af4da38a8585f3a0", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . X\n. O X\nO X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place O in row 1, column 2 (cell 2).", "cell_4": "Place O in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_2": 1.0, "cell_4": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.XXO.OXX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.X.OXOXO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.57, "cell_2": 0.43}, "solvable": {"false": 0.89, "true": 0.11}, "value": {"0": 0.43, "1": 0.44, "2": 0.13}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:e1d376dbb34a1efa3865f093", "state_id": "grid_navigation:e1d376dbb34a1efa3865f093", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # .\n# G A .\n# . . .\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "west": 1.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (0, 2), (1, 0), (2, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [1, 0], [2, 0]], "position": [1, 2], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.96, "east": 0.01, "south": 0.03}, "solvable": {"false": 0.12, "true": 0.88}, "value": {"0": 0.9, "1": 0.05, "2": 0.01, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:a5e5b760dee53c4d9839e8a5", "state_id": "tic_tac_toe:a5e5b760dee53c4d9839e8a5", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O O\nX O X\n. . X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_8": "Place O in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_7", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.5, "cell_8": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_7", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XXOXXOO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOOXOX..X", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.65, "cell_8": 0.35}, "solvable": {"false": 0.84, "true": 0.16}, "value": {"0": 0.42, "1": 0.43, "2": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:9c094a983a6305d4503d7357", "state_id": "tic_tac_toe:9c094a983a6305d4503d7357", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . .\nO . O\nX X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.5, "cell_5": 0.5, "cell_3": 0.0, "cell_2": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XO.O.XX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X..O.OXX.", "player": "O"}}}
{"id": "tic_tac_toe:4b56fcd095bf850a4f923dfb", "state_id": "tic_tac_toe:4b56fcd095bf850a4f923dfb", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X .\nX . O\nX . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place O in row 2, column 2 (cell 5).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 0.0, "cell_8": 0.0, "cell_9": 1.0, "cell_3": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XO.X.XO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OX.X.OX..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.14, "cell_5": 0.42, "cell_3": 0.14, "cell_8": 0.3}, "solvable": {"false": 0.88, "true": 0.12}, "value": {"0": 0.71, "1": 0.16, "2": 0.13}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:a81d3934f50dd276b8cbf584", "state_id": "grid_navigation:a81d3934f50dd276b8cbf584", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . #\nA . . #\n# . . G\n. . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (0, 3), (1, 0), (2, 3), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 3], [1, 3], [2, 0], [3, 2]], "position": [1, 0], "goal": [2, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.96, "north": 0.04}, "solvable": {"false": 0.16000000000000003, "true": 0.84}, "value": {"0": 0.01, "1": 0.76, "2": 0.17, "3": 0.06}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:cd510e5e38dced5d67e0f727", "state_id": "grid_navigation:cd510e5e38dced5d67e0f727", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # #\n. # . .\n. # # #\nA G . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 1.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 1), (1, 2), (1, 3), (2, 1), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [1, 1], [2, 1], [2, 2], [2, 3]], "position": [3, 0], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.63, "north": 0.37}, "solvable": {"false": 0.25, "true": 0.75}, "value": {"0": 0.93, "1": 0.03, "2": 0.02, "3": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:dc7f38712cee23ab29cfba1c", "state_id": "grid_navigation:dc7f38712cee23ab29cfba1c", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # #\n. # # #\n. G # .\n. A . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 1.0, "east": 0.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((1, 2), (1, 3), (2, 1), (2, 2), (2, 3), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [0, 3], [1, 1], [1, 2], [1, 3], [2, 2]], "position": [3, 1], "goal": [2, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.71, "east": 0.2, "west": 0.09}, "solvable": {"false": 0.41000000000000003, "true": 0.59}, "value": {"0": 0.18, "1": 0.39, "2": 0.18, "3": 0.25}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:dcd9bcf07c21978f3a14b17f", "state_id": "grid_navigation:dcd9bcf07c21978f3a14b17f", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # . #\n# G . #\n. . # .\n# A . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 1), (0, 3), (1, 0), (1, 3), (2, 2), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [0, 3], [1, 0], [1, 3], [2, 2], [3, 0]], "position": [3, 1], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.23, "north": 0.77}, "solvable": {"false": 0.29000000000000004, "true": 0.71}, "value": {"0": 0.08, "1": 0.6, "2": 0.2, "3": 0.12}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:cb1505cbe076756291906576", "state_id": "grid_navigation:cb1505cbe076756291906576", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . A\nG . . .\n. # . .\n. . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 0.5, "south": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[2, 1], [3, 3]], "position": [0, 3], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.13, "west": 0.87}, "solvable": {"false": 0.12, "true": 0.88}, "value": {"0": 0.11, "1": 0.75, "2": 0.12, "3": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:e42d9b1c42f848c7fe0af289", "state_id": "tic_tac_toe:e42d9b1c42f848c7fe0af289", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO O X\nX . .\n. O X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place X in row 2, column 2 (cell 5).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 0.5, "cell_6": 0.5, "cell_7": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXX..OOX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OOXX...OX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.61, "cell_6": 0.19, "cell_7": 0.2}, "solvable": {"false": 0.5900000000000001, "true": 0.41}, "value": {"0": 0.15, "1": 0.41, "2": 0.44}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:76904198caac2e5581534337", "state_id": "grid_navigation:76904198caac2e5581534337", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG . # #\n. . . .\n. # # .\n. . . A", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 2}, "gold_probs": {"action": {"north": 0.5, "west": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (0, 3), (2, 1), (2, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [2, 1], [2, 2]], "position": [3, 3], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.57, "north": 0.43}, "solvable": {"false": 0.31000000000000005, "true": 0.69}, "value": {"0": 0.01, "1": 0.2, "2": 0.64, "3": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:5c4c6c53b29c3cfc86bad5e1", "state_id": "tic_tac_toe:5c4c6c53b29c3cfc86bad5e1", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X .\nO . X\n. X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.0, "cell_3": 0.0, "cell_5": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOX.XOX.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OX.O.X.XO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.24, "cell_7": 0.27, "cell_5": 0.49}, "solvable": {"false": 0.5800000000000001, "true": 0.42}, "value": {"0": 0.12, "1": 0.4, "2": 0.48}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:8bdc3bb467aaba8954cd660b", "state_id": "grid_navigation:8bdc3bb467aaba8954cd660b", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. G # #\n# . A .\n. # . .\n. # . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "east": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (0, 3), (1, 0), (2, 1), (3, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [1, 0], [2, 1], [3, 1], [3, 3]], "position": [1, 2], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.83, "south": 0.06, "east": 0.11}, "solvable": {"false": 0.41000000000000003, "true": 0.59}, "value": {"0": 0.22, "1": 0.47, "2": 0.17, "3": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:a1422003ce1020dfb730e9e7", "state_id": "grid_navigation:a1422003ce1020dfb730e9e7", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . #\n# . . .\n# . . .\n. A # G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 1.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (1, 3), (2, 3), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 0], [2, 0], [3, 2]], "position": [3, 1], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.87, "west": 0.13}, "solvable": {"false": 0.56, "true": 0.44}, "value": {"0": 0.23, "1": 0.29, "2": 0.17, "3": 0.31}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:477a808736e0368ec5766281", "state_id": "grid_navigation:477a808736e0368ec5766281", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nA . . #\n. . . G\n. . . #\n# # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.5, "south": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 2), (1, 3), (2, 3), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [2, 3], [3, 0], [3, 1], [3, 2]], "position": [0, 0], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.7, "south": 0.3}, "solvable": {"false": 0.07999999999999996, "true": 0.92}, "value": {"0": 0.04, "1": 0.78, "2": 0.14, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:be46bde9af0dccd519f484ef", "state_id": "grid_navigation:be46bde9af0dccd519f484ef", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . #\n. . # G\n# . . .\n. A # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 1.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (1, 1), (1, 3), (2, 0), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 3], [1, 2], [2, 0], [3, 2]], "position": [3, 1], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.31, "north": 0.69}, "solvable": {"false": 0.31999999999999995, "true": 0.68}, "value": {"0": 0.03, "1": 0.54, "2": 0.27, "3": 0.16}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:27a234b157042962cc526e5a", "state_id": "tic_tac_toe:27a234b157042962cc526e5a", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . O\nX X .\n. . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.0, "cell_2": 0.5, "cell_7": 0.0, "cell_8": 0.0, "cell_6": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....XXO.O:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.OXX....", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.54, "cell_7": 0.1, "cell_8": 0.07, "cell_9": 0.05, "cell_2": 0.24}, "solvable": {"false": 0.5, "true": 0.5}, "value": {"0": 0.18, "1": 0.28, "2": 0.54}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:0f636a9313798bb0e04d58a5", "state_id": "tic_tac_toe:0f636a9313798bb0e04d58a5", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X X\nX . O\nO X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place O in row 2, column 2 (cell 5).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_5": 1.0, "cell_9": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXX.XOXO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXXX.OOX.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.5, "cell_9": 0.5}, "solvable": {"false": 0.86, "true": 0.14}, "value": {"0": 0.43, "1": 0.4, "2": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:f6318934bc040e6a3b60f9f5", "state_id": "tic_tac_toe:f6318934bc040e6a3b60f9f5", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X O\n. . O\nX . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.5, "cell_4": 0.5, "cell_8": 0.0, "cell_5": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XO..OXX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XXO..OX..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.36, "cell_4": 0.27, "cell_9": 0.16, "cell_8": 0.21}, "solvable": {"false": 0.83, "true": 0.17}, "value": {"0": 0.42, "1": 0.36, "2": 0.22}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:2817434fedff2c3469d128d2", "state_id": "grid_navigation:2817434fedff2c3469d128d2", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . G\n. . A #\n# . . #\n. # . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "south": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 2), (0, 3), (2, 0), (2, 3), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 3], [2, 0], [2, 3], [3, 1], [3, 3]], "position": [1, 2], "goal": [0, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.2, "north": 0.76, "south": 0.04}, "solvable": {"false": 0.30000000000000004, "true": 0.7}, "value": {"0": 0.4, "1": 0.39, "2": 0.1, "3": 0.11}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:17f280299784e38fa0ef51e8", "state_id": "tic_tac_toe:17f280299784e38fa0ef51e8", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . X\nO O X\n. X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place O in row 1, column 2 (cell 2).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_1": "Place O in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_2": 0.0, "cell_9": 1.0, "cell_7": 0.0, "cell_1": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOOX.X.:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..XOOX.X.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.19, "cell_1": 0.26, "cell_9": 0.25, "cell_7": 0.3}, "solvable": {"false": 0.85, "true": 0.15}, "value": {"0": 0.48, "1": 0.34, "2": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:2e80703dda5d568187566b1c", "state_id": "grid_navigation:2e80703dda5d568187566b1c", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A # .\n. G # #\n. # . .\n# . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 1.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 2), (1, 2), (1, 3), (2, 1), (3, 0), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 2], [1, 3], [2, 1], [3, 0], [3, 2], [3, 3]], "position": [0, 1], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.64, "south": 0.36}, "solvable": {"false": 0.4, "true": 0.6}, "value": {"0": 0.61, "1": 0.17, "2": 0.09, "3": 0.13}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:3cad7406d6dbf8ce2f854d1b", "state_id": "tic_tac_toe:3cad7406d6dbf8ce2f854d1b", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O O\nX . .\n. O X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place X in row 2, column 3 (cell 6).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.0, "cell_7": 0.5, "cell_5": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXX..XOO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOOX...OX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.57, "cell_6": 0.21, "cell_7": 0.22}, "solvable": {"false": 0.64, "true": 0.36}, "value": {"0": 0.21, "1": 0.43, "2": 0.36}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:678ff3a024ea7ba3a601ed49", "state_id": "grid_navigation:678ff3a024ea7ba3a601ed49", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A G .\n. . . #\n. # . .\n. # . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 1.0, "south": 0.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (2, 2), (3, 0), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 3], [2, 1], [3, 1], [3, 3]], "position": [0, 1], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.01, "west": 0.01, "east": 0.98}, "solvable": {"false": 0.040000000000000036, "true": 0.96}, "value": {"0": 0.99, "1": 0.01, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:766c4a01390510d425e9b11c", "state_id": "tic_tac_toe:766c4a01390510d425e9b11c", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O .\n. X O\n. X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place X in row 1, column 3 (cell 3).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_3": 1.0, "cell_4": 0.0, "cell_7": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XXXOOO.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XO..XO.XO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.45, "cell_3": 0.31, "cell_4": 0.24}, "solvable": {"false": 0.62, "true": 0.38}, "value": {"0": 0.4, "1": 0.21, "2": 0.39}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:9b05de82e1107e77c3cd48ee", "state_id": "grid_navigation:9b05de82e1107e77c3cd48ee", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . .\n. . . .\nA # # .\n. # . G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 2}, "gold_probs": {"action": {"north": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (1, 1), (1, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [2, 1], [2, 2], [3, 1]], "position": [2, 0], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.62, "south": 0.38}, "solvable": {"false": 0.41000000000000003, "true": 0.59}, "value": {"0": 0.01, "1": 0.36, "2": 0.5, "3": 0.13}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:b2d5d5cad01d4ce27eacbea3", "state_id": "grid_navigation:b2d5d5cad01d4ce27eacbea3", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . A .\n. . . #\n. # # G\n# . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"south": 0.3333333333333333, "west": 0.3333333333333333, "east": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 1), (2, 1), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 3], [2, 1], [2, 2], [3, 0]], "position": [0, 2], "goal": [2, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.5, "west": 0.16, "east": 0.34}, "solvable": {"false": 0.38, "true": 0.62}, "value": {"0": 0.01, "1": 0.4, "2": 0.41, "3": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:58df0a5a4d5ba8e4593e8b84", "state_id": "tic_tac_toe:58df0a5a4d5ba8e4593e8b84", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X O\nO . X\nX . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place O in row 3, column 2 (cell 8).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_1": "Place O in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_8": 0.3333333333333333, "cell_9": 0.3333333333333333, "cell_5": 0.3333333333333333, "cell_1": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5", "cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XX.OOX.:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XOO.XX..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.45, "cell_1": 0.11, "cell_8": 0.26, "cell_9": 0.18}, "solvable": {"false": 0.85, "true": 0.15}, "value": {"0": 0.31, "1": 0.43, "2": 0.26}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:83a92fe2593ef8100eaa0e4d", "state_id": "tic_tac_toe:83a92fe2593ef8100eaa0e4d", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X O\n. X O\nX . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.0, "cell_8": 0.0, "cell_4": 0.0, "cell_9": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOX.OX.:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XO.XOX..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.42, "cell_4": 0.2, "cell_9": 0.17, "cell_1": 0.21}, "solvable": {"false": 0.81, "true": 0.19}, "value": {"0": 0.53, "1": 0.26, "2": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:12566ecdfce08624a4fec4db", "state_id": "grid_navigation:12566ecdfce08624a4fec4db", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\n# . # .\n# # . A\n. G . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 0), (1, 1), (2, 0), (2, 2), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 0], [1, 2], [2, 0], [2, 1], [3, 3]], "position": [2, 3], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.75, "north": 0.25}, "solvable": {"false": 0.33999999999999997, "true": 0.66}, "value": {"0": 0.09, "1": 0.62, "2": 0.17, "3": 0.12}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:3def007cdeb027ba65050952", "state_id": "grid_navigation:3def007cdeb027ba65050952", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # #\n. # . G\n. . # #\n# . A .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.5, "east": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 2), (1, 0), (1, 2), (2, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [1, 1], [2, 2], [2, 3], [3, 0]], "position": [3, 2], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.47, "west": 0.53}, "solvable": {"false": 0.44999999999999996, "true": 0.55}, "value": {"0": 0.02, "1": 0.44, "2": 0.31, "3": 0.23}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:9847543bf41bee7d7526c19b", "state_id": "grid_navigation:9847543bf41bee7d7526c19b", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . #\n. # . A\n# # # G\n# . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 1.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 0), (1, 1), (2, 1), (2, 2), (2, 3), (3, 0), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 3], [1, 1], [2, 0], [2, 1], [2, 2], [3, 0], [3, 2]], "position": [1, 3], "goal": [2, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.5, "west": 0.5}, "solvable": {"false": 0.69, "true": 0.31}, "value": {"0": 0.21, "1": 0.22, "2": 0.12, "3": 0.45}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:0169facd35cd593b531f5309", "state_id": "grid_navigation:0169facd35cd593b531f5309", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . G #\n# . # .\n. # . .\nA . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (1, 1), (1, 3), (2, 2), (3, 0), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 0], [1, 2], [2, 1], [3, 2], [3, 3]], "position": [3, 0], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.57, "east": 0.43}, "solvable": {"false": 0.38, "true": 0.62}, "value": {"0": 0.02, "1": 0.42, "2": 0.4, "3": 0.16}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:147c60c2cc08d4f48c202531", "state_id": "grid_navigation:147c60c2cc08d4f48c202531", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . A\n. # . .\n. G . .\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 0.5, "south": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((1, 2),)", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 1]], "position": [0, 3], "goal": [2, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.34, "south": 0.66}, "solvable": {"false": 0.13, "true": 0.87}, "value": {"0": 0.05, "1": 0.81, "2": 0.12, "3": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:7b0f70e4f45b33c5f108e2a7", "state_id": "tic_tac_toe:7b0f70e4f45b33c5f108e2a7", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O X\nO O X\n. X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_7": 0.0, "cell_9": 1.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXXOO.XX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOXOOX.X.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.67, "cell_9": 0.33}, "solvable": {"false": 0.83, "true": 0.17}, "value": {"0": 0.33, "1": 0.41000000000000003, "2": 0.26}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:d08d545c158238d154223e0f", "state_id": "tic_tac_toe:d08d545c158238d154223e0f", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . X\nO X O\n. O X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_1": "Place X in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.3333333333333333, "cell_2": 0.3333333333333333, "cell_1": 0.3333333333333333}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_2", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOXO.OX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..XOXO.OX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.41, "cell_2": 0.15, "cell_1": 0.44}, "solvable": {"false": 0.62, "true": 0.38}, "value": {"0": 0.07, "1": 0.35, "2": 0.58}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:8b1fc28efd912ee77206ee1e", "state_id": "grid_navigation:8b1fc28efd912ee77206ee1e", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # #\n. # . .\n. # # G\nA . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 0), (1, 1), (1, 3), (2, 1), (2, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [0, 3], [1, 1], [2, 1], [2, 2], [3, 2]], "position": [3, 0], "goal": [2, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.63, "north": 0.37}, "solvable": {"false": 0.54, "true": 0.46}, "value": {"0": 0.01, "1": 0.35, "2": 0.36, "3": 0.28}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:6b90610ec1eef4ba499e6bd8", "state_id": "grid_navigation:6b90610ec1eef4ba499e6bd8", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # .\n# . # .\n# A . G\n. # # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 1.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (1, 0), (1, 2), (1, 3), (2, 0), (3, 1), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [1, 0], [1, 2], [2, 0], [3, 1], [3, 2], [3, 3]], "position": [2, 1], "goal": [2, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.94, "north": 0.06}, "solvable": {"false": 0.10999999999999999, "true": 0.89}, "value": {"0": 0.79, "1": 0.11, "2": 0.04, "3": 0.06}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:f8aedaff85af986f280f9d2e", "state_id": "grid_navigation:f8aedaff85af986f280f9d2e", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . A\n. . G .\n. # # .\n# . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.5, "west": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 3), (2, 1), (2, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [2, 1], [2, 2], [3, 0]], "position": [0, 3], "goal": [1, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.69, "west": 0.31}, "solvable": {"false": 0.10999999999999999, "true": 0.89}, "value": {"0": 0.94, "1": 0.04, "2": 0.01, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:5d7ba4e898eae5710d9e2f23", "state_id": "grid_navigation:5d7ba4e898eae5710d9e2f23", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG . # #\n# # . A\n. # . .\n# . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": false, "value": 3}, "gold_probs": {"action": {"south": 0.5, "west": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 3), (1, 1), (1, 2), (2, 0), (3, 0), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [1, 0], [1, 1], [2, 1], [3, 0], [3, 3]], "position": [1, 3], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.31, "west": 0.69}, "solvable": {"false": 0.5, "true": 0.5}, "value": {"0": 0.1, "1": 0.42, "2": 0.26, "3": 0.22}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:3d2b5ea67d3ad5115f8e18a7", "state_id": "grid_navigation:3d2b5ea67d3ad5115f8e18a7", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG # . .\n# . A .\n# . # #\n# # . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.3333333333333333, "east": 0.3333333333333333, "north": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 2), (0, 3), (1, 0), (1, 3), (2, 2), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 0], [2, 0], [2, 2], [2, 3], [3, 0], [3, 1], [3, 3]], "position": [1, 2], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.55, "west": 0.23, "east": 0.22}, "solvable": {"false": 0.56, "true": 0.44}, "value": {"0": 0.2, "1": 0.42, "2": 0.13, "3": 0.25}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:2ed1dc7b555b81bc252136f6", "state_id": "grid_navigation:2ed1dc7b555b81bc252136f6", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. G . .\n. . # #\n# # A .\n. # . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "south": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 2), (1, 3), (2, 0), (2, 1), (3, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 2], [1, 3], [2, 0], [2, 1], [3, 1], [3, 3]], "position": [2, 2], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.75, "south": 0.25}, "solvable": {"false": 0.39, "true": 0.61}, "value": {"0": 0.03, "1": 0.5, "2": 0.32, "3": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:db8767d45ffc45920deeb239", "state_id": "tic_tac_toe:db8767d45ffc45920deeb239", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X O\n. X .\n. O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place O in row 2, column 3 (cell 6).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_1": "Place O in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_6": 0.0, "cell_4": 0.0, "cell_7": 0.0, "cell_1": 1.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...OXXX.O:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XO.X..OX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.29, "cell_4": 0.22, "cell_1": 0.29, "cell_7": 0.2}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.38, "1": 0.43, "2": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:78f293f48dc4113c164222c4", "state_id": "grid_navigation:78f293f48dc4113c164222c4", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # #\n. G . #\n. A # #\n# # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 1.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 2), (0, 3), (1, 3), (2, 2), (2, 3), (3, 0), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [1, 3], [2, 2], [2, 3], [3, 0], [3, 1]], "position": [2, 1], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.88, "west": 0.12}, "solvable": {"false": 0.27, "true": 0.73}, "value": {"0": 0.7, "1": 0.14, "2": 0.06, "3": 0.1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:934c2fc20e199714c433d8e6", "state_id": "grid_navigation:934c2fc20e199714c433d8e6", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # G #\n# . # .\n# . . .\n# . A .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.3333333333333333, "west": 0.3333333333333333, "east": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 2), (1, 1), (1, 3), (2, 3), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 3], [1, 0], [1, 2], [2, 0], [3, 0]], "position": [3, 2], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.79, "east": 0.06, "west": 0.15}, "solvable": {"false": 0.49, "true": 0.51}, "value": {"0": 0.01, "1": 0.36, "2": 0.41, "3": 0.22}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:c9fab57bdbed4d0374cc8c9c", "state_id": "tic_tac_toe:c9fab57bdbed4d0374cc8c9c", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O X\n. X .\n. O O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place X in row 2, column 3 (cell 6).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_7", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.0, "cell_4": 0.0, "cell_7": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOXOO.X:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOX.X..OO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.25, "cell_7": 0.31, "cell_4": 0.44}, "solvable": {"false": 0.65, "true": 0.35}, "value": {"0": 0.2, "1": 0.38, "2": 0.42}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:bb05ddd3b99da40c069b7d86", "state_id": "tic_tac_toe:bb05ddd3b99da40c069b7d86", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . .\nO X X\nO O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_3": "Place X in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 1.0, "cell_2": 0.0, "cell_3": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XXXO.OO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X..OXXOO.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.26, "cell_2": 0.13, "cell_9": 0.61}, "solvable": {"false": 0.36, "true": 0.64}, "value": {"0": 0.19, "1": 0.19, "2": 0.62}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:14411800c43cac7195a5adaf", "state_id": "tic_tac_toe:14411800c43cac7195a5adaf", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\n. . O\nX O X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_1": "Place X in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_2": 0.0, "cell_3": 0.0, "cell_5": 0.5, "cell_1": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.....OXOX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".....OXOX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.12, "cell_1": 0.12, "cell_2": 0.16, "cell_5": 0.41000000000000003, "cell_4": 0.19}, "solvable": {"false": 0.6799999999999999, "true": 0.32}, "value": {"0": 0.26, "1": 0.42, "2": 0.32}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:feecae11ef03a9ede6983076", "state_id": "tic_tac_toe:feecae11ef03a9ede6983076", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O O\n. . .\n. . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_4": "Place X in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.25, "cell_8": 0.25, "cell_5": 0.25, "cell_6": 0.0, "cell_4": 0.25}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4", "cell_5", "cell_7", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..X...XOO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOO.....X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.07, "cell_5": 0.56, "cell_4": 0.14, "cell_8": 0.09, "cell_7": 0.14}, "solvable": {"false": 0.8200000000000001, "true": 0.18}, "value": {"0": 0.21, "1": 0.54, "2": 0.25}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:f82eb41929099402d071fe3f", "state_id": "tic_tac_toe:f82eb41929099402d071fe3f", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . O\nO . X\n. . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place X in row 2, column 2 (cell 5).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_8": "Place X in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_5": 0.0, "cell_1": 0.5, "cell_2": 0.0, "cell_7": 0.5, "cell_8": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OO.X..X:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..OO.X..X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.13, "cell_5": 0.51, "cell_2": 0.12, "cell_7": 0.11, "cell_8": 0.13}, "solvable": {"false": 0.6799999999999999, "true": 0.32}, "value": {"0": 0.2, "1": 0.43, "2": 0.37}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:ca3a305a09df81aa34fec468", "state_id": "tic_tac_toe:ca3a305a09df81aa34fec468", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X O\nO X .\nX . X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place O in row 2, column 3 (cell 6).", "cell_8": "Place O in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_8", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_6": 0.0, "cell_8": 1.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.XXX.OOX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXOOX.X.X", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.51, "cell_6": 0.49}, "solvable": {"false": 0.83, "true": 0.17}, "value": {"0": 0.45, "1": 0.37, "2": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:f9a9bd9480b83869217cc7b9", "state_id": "grid_navigation:f9a9bd9480b83869217cc7b9", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # G\n. # # .\n. A . #\n. . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.3333333333333333, "south": 0.3333333333333333, "west": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (1, 1), (1, 2), (2, 0), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 1], [1, 2], [2, 3], [3, 2]], "position": [2, 1], "goal": [0, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.35, "south": 0.16, "east": 0.49}, "solvable": {"false": 0.6699999999999999, "true": 0.33}, "value": {"0": 0.02, "1": 0.33, "2": 0.38, "3": 0.27}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:d1a22e21f9fa5ecfd2cc6c0b", "state_id": "tic_tac_toe:d1a22e21f9fa5ecfd2cc6c0b", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X X\nO X .\n. . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place O in row 2, column 3 (cell 6).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_8": "Place O in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_7", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.0, "cell_9": 0.0, "cell_7": 1.0, "cell_8": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....XOXXO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXXOX....", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.1, "cell_9": 0.18, "cell_6": 0.59, "cell_7": 0.13}, "solvable": {"false": 0.92, "true": 0.08}, "value": {"0": 0.62, "1": 0.24, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:cbe9297ead069fa76574a670", "state_id": "grid_navigation:cbe9297ead069fa76574a670", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. G . #\n# # # .\n. A . #\n# . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.3333333333333333, "west": 0.3333333333333333, "south": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 0), (1, 1), (1, 2), (2, 3), (3, 0), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 0], [1, 1], [1, 2], [2, 3], [3, 0], [3, 2]], "position": [2, 1], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.42, "south": 0.06, "west": 0.52}, "solvable": {"false": 0.41000000000000003, "true": 0.59}, "value": {"0": 0.17, "1": 0.45, "2": 0.18, "3": 0.2}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:b09e4a16b82eed09ca7c0838", "state_id": "grid_navigation:b09e4a16b82eed09ca7c0838", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # #\n. . G #\n. . A #\n. . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "west": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 1), (0, 2), (0, 3), (1, 0), (2, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [0, 3], [1, 3], [2, 3], [3, 3]], "position": [2, 2], "goal": [1, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.06, "south": 0.02, "north": 0.92}, "solvable": {"false": 0.08999999999999997, "true": 0.91}, "value": {"0": 0.95, "1": 0.04, "2": 0, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:460d65643edf650069454f5e", "state_id": "grid_navigation:460d65643edf650069454f5e", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . A\nG # # .\n# # . #\n# # # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": false, "value": 3}, "gold_probs": {"action": {"south": 0.5, "west": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (0, 3), (1, 0), (1, 1), (1, 2), (1, 3), (2, 1), (2, 3), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 1], [1, 2], [2, 0], [2, 1], [2, 3], [3, 0], [3, 1], [3, 2], [3, 3]], "position": [0, 3], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.42, "west": 0.58}, "solvable": {"false": 0.6599999999999999, "true": 0.34}, "value": {"0": 0.06, "1": 0.32, "2": 0.16, "3": 0.46}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:fefa7c4f9c8a522319f45361", "state_id": "grid_navigation:fefa7c4f9c8a522319f45361", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # G\n. A # #\n# . . .\n. . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.3333333333333333, "west": 0.3333333333333333, "south": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 3), (1, 0), (1, 1), (2, 3), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [1, 2], [1, 3], [2, 0], [3, 3]], "position": [1, 1], "goal": [0, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.71, "south": 0.09, "west": 0.2}, "solvable": {"false": 0.5700000000000001, "true": 0.43}, "value": {"0": 0.3, "1": 0.28, "2": 0.14, "3": 0.28}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:f4f8b04bb83350a73ff94071", "state_id": "grid_navigation:f4f8b04bb83350a73ff94071", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # #\n. A # .\n. . G .\n# . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "west": 0.0, "south": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 3), (1, 2), (1, 3), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [1, 2], [3, 0], [3, 3]], "position": [1, 1], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.02, "south": 0.82, "west": 0.16}, "solvable": {"false": 0.10999999999999999, "true": 0.89}, "value": {"0": 0.76, "1": 0.19, "2": 0.02, "3": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:c964671554d0d66b03e11645", "state_id": "grid_navigation:c964671554d0d66b03e11645", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG # . .\n. . . .\n. . . A\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.5, "south": 0.0, "west": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1),)", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1]], "position": [2, 3], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.01, "north": 0.56, "west": 0.43}, "solvable": {"false": 0.17000000000000004, "true": 0.83}, "value": {"0": 0.01, "1": 0.52, "2": 0.4, "3": 0.07}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:e8103c3eb601631e485f0852", "state_id": "grid_navigation:e8103c3eb601631e485f0852", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # G .\n. . . #\n# . # .\n# A . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (0, 3), (1, 0), (2, 1), (2, 3), (3, 0), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [1, 3], [2, 0], [2, 2], [3, 0], [3, 3]], "position": [3, 1], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.76, "east": 0.24}, "solvable": {"false": 0.33999999999999997, "true": 0.66}, "value": {"0": 0.02, "1": 0.46, "2": 0.38, "3": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:2dddc4f0218dc43ca58c5378", "state_id": "grid_navigation:2dddc4f0218dc43ca58c5378", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n. G . .\nA . . #\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.5, "east": 0.5, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 2), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [2, 3]], "position": [2, 0], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.25, "south": 0.01, "north": 0.74}, "solvable": {"false": 0.06000000000000005, "true": 0.94}, "value": {"0": 0.5, "1": 0.46, "2": 0.03, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:7d343f42132859760ec0f543", "state_id": "grid_navigation:7d343f42132859760ec0f543", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . A\n. # # .\n. G # #\n# . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": false, "value": 3}, "gold_probs": {"action": {"south": 0.5, "west": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (1, 2), (1, 3), (2, 1), (2, 2), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 1], [1, 2], [2, 2], [2, 3], [3, 0]], "position": [0, 3], "goal": [2, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.55, "west": 0.45}, "solvable": {"false": 0.54, "true": 0.46}, "value": {"0": 0.02, "1": 0.31, "2": 0.42, "3": 0.25}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:d6e710ede272f3a8d3b98067", "state_id": "tic_tac_toe:d6e710ede272f3a8d3b98067", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O X\nO O .\n. X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_6", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_9": 0.0, "cell_6": 1.0, "cell_7": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XXOO.OX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOXOO..X.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.22, "cell_7": 0.2, "cell_6": 0.58}, "solvable": {"false": 0.6599999999999999, "true": 0.34}, "value": {"0": 0.37, "1": 0.36, "2": 0.27}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:079227fc0b85ecf67f564bff", "state_id": "grid_navigation:079227fc0b85ecf67f564bff", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . #\nG . . #\n# . A .\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east.", "west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.0, "east": 0.0, "west": 0.5, "north": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (3, 0), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 3], [2, 0]], "position": [2, 2], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.05, "east": 0.01, "west": 0.65, "north": 0.29}, "solvable": {"false": 0.15000000000000002, "true": 0.85}, "value": {"0": 0.03, "1": 0.63, "2": 0.27, "3": 0.07}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:65d1d27cd132310f53c00a1d", "state_id": "grid_navigation:65d1d27cd132310f53c00a1d", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . G .\n. . . .\nA . # #\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.0, "east": 0.5, "north": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((2, 0), (2, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[2, 2], [2, 3]], "position": [2, 0], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.9299999999999999, "south": 0.03, "east": 0.04}, "solvable": {"false": 0.06999999999999995, "true": 0.93}, "value": {"0": 0.01, "1": 0.74, "2": 0.23, "3": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:dfdd34252f4b5cb1306a6a06", "state_id": "grid_navigation:dfdd34252f4b5cb1306a6a06", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . A G\n. . . .\n. . # .\n# . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "east": 1.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 3), (2, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [2, 2], [3, 0]], "position": [0, 2], "goal": [0, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.98, "south": 0.01, "west": 0.01}, "solvable": {"false": 0.07999999999999996, "true": 0.92}, "value": {"0": 0.97, "1": 0.01, "2": 0.01, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:8000c70ea178d8dccac9e3cb", "state_id": "tic_tac_toe:8000c70ea178d8dccac9e3cb", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O X\n. . O\nO . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_8": "Place X in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_5": 1.0, "cell_8": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.X..OXOX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOX..OO.X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.29, "cell_4": 0.21, "cell_5": 0.5}, "solvable": {"false": 0.72, "true": 0.28}, "value": {"0": 0.22, "1": 0.47, "2": 0.31}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:aa01bfd864f36587cccba95b", "state_id": "tic_tac_toe:aa01bfd864f36587cccba95b", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . X\nX O O\n. X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_1": "Place O in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_9": 0.0, "cell_2": 0.0, "cell_7": 0.5, "cell_1": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XXOO.X.:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..XXOO.X.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.21, "cell_2": 0.26, "cell_7": 0.26, "cell_1": 0.27}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.37, "1": 0.44, "2": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:8889172d81fe4186293fc5d0", "state_id": "tic_tac_toe:8889172d81fe4186293fc5d0", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . O\n. X .\nX . O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_8": "Place X in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_6", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_4": 0.0, "cell_2": 0.0, "cell_1": 0.0, "cell_6": 1.0, "cell_8": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O.X.X.O:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..O.X.X.O", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.28, "cell_6": 0.14, "cell_1": 0.18, "cell_4": 0.23, "cell_2": 0.17}, "solvable": {"false": 0.77, "true": 0.23}, "value": {"0": 0.22, "1": 0.43, "2": 0.35}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:54a35113155f2e9f4e69a8fb", "state_id": "tic_tac_toe:54a35113155f2e9f4e69a8fb", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . X\nO X .\n. X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_6": "Place O in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_7", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 1.0, "cell_2": 0.0, "cell_9": 0.0, "cell_6": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XXX..OO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.XOX..X.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.29, "cell_2": 0.1, "cell_7": 0.35, "cell_9": 0.26}, "solvable": {"false": 0.89, "true": 0.11}, "value": {"0": 0.5, "1": 0.35, "2": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:94c8adc590dee4a25d2d3c22", "state_id": "tic_tac_toe:94c8adc590dee4a25d2d3c22", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O O\n. . X\n. X X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_1": "Place O in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.5, "cell_5": 0.0, "cell_4": 0.0, "cell_1": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...O.XOXX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".OO..X.XX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.21, "cell_1": 0.25, "cell_7": 0.28, "cell_5": 0.26}, "solvable": {"false": 0.8200000000000001, "true": 0.18}, "value": {"0": 0.49, "1": 0.32, "2": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:c53c527bcf4e458749edae53", "state_id": "grid_navigation:c53c527bcf4e458749edae53", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n# . # #\n. . G .\n. # . A", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.5, "west": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 2), (1, 2), (2, 0), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [1, 2], [1, 3], [3, 1]], "position": [3, 3], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.28, "north": 0.72}, "solvable": {"false": 0.20999999999999996, "true": 0.79}, "value": {"0": 0.3, "1": 0.54, "2": 0.11, "3": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:a4e09484420068267b151ba5", "state_id": "tic_tac_toe:a4e09484420068267b151ba5", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO O .\nX X O\n. X X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place O in row 1, column 3 (cell 3).", "cell_7": "Place O in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_3": 1.0, "cell_7": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOOXXXX.:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OO.XXO.XX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.5, "cell_7": 0.5}, "solvable": {"false": 0.8200000000000001, "true": 0.18}, "value": {"0": 0.49, "1": 0.32, "2": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:92d0d7802b9a56abef2c8d8a", "state_id": "tic_tac_toe:92d0d7802b9a56abef2c8d8a", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\nO . X\nX O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place O in row 2, column 2 (cell 5).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_1": "Place O in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_5": 0.0, "cell_2": 0.0, "cell_3": 1.0, "cell_1": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...O.XXOX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "...O.XXOX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.16, "cell_1": 0.17, "cell_2": 0.24, "cell_5": 0.43}, "solvable": {"false": 0.9, "true": 0.1}, "value": {"0": 0.46, "1": 0.37, "2": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:d070304b4664aa570c27c2bd", "state_id": "tic_tac_toe:d070304b4664aa570c27c2bd", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O O\n. X .\n. . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place X in row 3, column 2 (cell 8).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_1": "Place X in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_8": 0.0, "cell_7": 0.0, "cell_4": 0.0, "cell_6": 0.0, "cell_1": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....XOX.O:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".OO.X...X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.16, "cell_4": 0.19, "cell_6": 0.19, "cell_7": 0.2, "cell_1": 0.26}, "solvable": {"false": 0.6599999999999999, "true": 0.34}, "value": {"0": 0.22, "1": 0.36, "2": 0.42}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:841d9407e6625c67e6633f30", "state_id": "grid_navigation:841d9407e6625c67e6633f30", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # .\n. . . #\n. # G .\n# . A .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 1.0, "west": 0.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 2), (1, 3), (2, 1), (2, 3), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [1, 3], [2, 1], [3, 0]], "position": [3, 2], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.13, "north": 0.57, "west": 0.3}, "solvable": {"false": 0.22999999999999998, "true": 0.77}, "value": {"0": 0.15, "1": 0.62, "2": 0.16, "3": 0.07}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:1172305bfe5ff18b042dc66d", "state_id": "tic_tac_toe:1172305bfe5ff18b042dc66d", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . O\n. . .\nX . O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place X in row 3, column 2 (cell 8).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_8": 0.0, "cell_6": 0.0, "cell_4": 1.0, "cell_5": 0.0, "cell_2": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.O...X.X:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.O...X.O", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.18, "cell_8": 0.1, "cell_2": 0.12, "cell_5": 0.53, "cell_6": 0.07}, "solvable": {"false": 0.78, "true": 0.22}, "value": {"0": 0.23, "1": 0.52, "2": 0.25}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:58686404db8285254c01d1b9", "state_id": "grid_navigation:58686404db8285254c01d1b9", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n. G . #\nA . . .\n. # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "north": 0.5, "east": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 2), (1, 3), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 3], [3, 1]], "position": [2, 0], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.33, "north": 0.66, "south": 0.01}, "solvable": {"false": 0.08999999999999997, "true": 0.91}, "value": {"0": 0.43, "1": 0.51, "2": 0.05, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:eafa0c1d9ff8988f52b1df96", "state_id": "grid_navigation:eafa0c1d9ff8988f52b1df96", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\n. . . A\n# . . G\n. . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "south": 1.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (2, 3), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [2, 0], [3, 3]], "position": [1, 3], "goal": [2, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.05, "west": 0.19, "south": 0.76}, "solvable": {"false": 0.09999999999999998, "true": 0.9}, "value": {"0": 0.77, "1": 0.2, "2": 0.02, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:c4d039b79ffe5493ed61e16f", "state_id": "tic_tac_toe:c4d039b79ffe5493ed61e16f", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . X\nX O .\nO O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place X in row 1, column 2 (cell 2).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 1.0, "cell_6": 0.0, "cell_9": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOO.OXX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.XXO.OO.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.21, "cell_6": 0.34, "cell_9": 0.45}, "solvable": {"false": 0.44999999999999996, "true": 0.55}, "value": {"0": 0.19, "1": 0.25, "2": 0.56}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:bf757b4bc6d0a49c59243bac", "state_id": "grid_navigation:bf757b4bc6d0a49c59243bac", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # G .\n. # . .\n. . # .\n. A . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 2}, "gold_probs": {"action": {"north": 0.0, "east": 1.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 2), (2, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 1], [2, 2]], "position": [3, 1], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.12, "north": 0.58, "east": 0.3}, "solvable": {"false": 0.48, "true": 0.52}, "value": {"0": 0.01, "1": 0.27, "2": 0.51, "3": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:6bf771ab5ecf9ef20a8033c2", "state_id": "tic_tac_toe:6bf771ab5ecf9ef20a8033c2", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO O .\nX . X\n. . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.0, "cell_7": 0.0, "cell_8": 0.0, "cell_3": 0.5, "cell_5": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...X.X.OO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OO.X.X...", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.15, "cell_9": 0.06, "cell_3": 0.17, "cell_8": 0.1, "cell_5": 0.52}, "solvable": {"false": 0.64, "true": 0.36}, "value": {"0": 0.22, "1": 0.37, "2": 0.41}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:1cb2156bd490e2f55a3fd10e", "state_id": "tic_tac_toe:1cb2156bd490e2f55a3fd10e", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O O\n. . X\nX . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 1.0, "cell_4": 0.0, "cell_8": 0.0, "cell_5": 0.0, "cell_9": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XO..OX.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".OO..XX..", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.16, "cell_1": 0.23, "cell_4": 0.18, "cell_5": 0.33, "cell_9": 0.1}, "solvable": {"false": 0.71, "true": 0.29}, "value": {"0": 0.25, "1": 0.45, "2": 0.3}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:4ffe5e93e8120039870b44ad", "state_id": "grid_navigation:4ffe5e93e8120039870b44ad", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n# A G .\n. # . .\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 1.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (1, 3), (2, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 0], [2, 1]], "position": [1, 1], "goal": [1, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0, "east": 1}, "solvable": {"false": 0.040000000000000036, "true": 0.96}, "value": {"0": 0.99, "1": 0, "2": 0, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:0957cd8162e9de42c97e4ce5", "state_id": "tic_tac_toe:0957cd8162e9de42c97e4ce5", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . O\nO X .\n. X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.0, "cell_6": 0.0, "cell_2": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXXX.O.O:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.OOX..XO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.38, "cell_2": 0.24, "cell_6": 0.38}, "solvable": {"false": 0.7, "true": 0.3}, "value": {"0": 0.15, "1": 0.44, "2": 0.41}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:4d3c7a406f7c6865b77ec4a3", "state_id": "grid_navigation:4d3c7a406f7c6865b77ec4a3", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. # . #\n# G # .\n. A . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.0, "north": 1.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (1, 2), (2, 1), (3, 0), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 1], [1, 3], [2, 0], [2, 2], [3, 3]], "position": [3, 1], "goal": [2, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.19, "east": 0.14, "north": 0.67}, "solvable": {"false": 0.44999999999999996, "true": 0.55}, "value": {"0": 0.12, "1": 0.5, "2": 0.2, "3": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:644fd8fc93c38d0d215f70bb", "state_id": "grid_navigation:644fd8fc93c38d0d215f70bb", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG # # .\n. . # #\n. # . .\nA . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 1.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 2), (1, 2), (1, 3), (2, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [1, 2], [1, 3], [2, 1], [3, 3]], "position": [3, 0], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.8, "east": 0.2}, "solvable": {"false": 0.48, "true": 0.52}, "value": {"0": 0.01, "1": 0.3, "2": 0.36, "3": 0.33}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:b819a63d609ef60f31d74a79", "state_id": "grid_navigation:b819a63d609ef60f31d74a79", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . #\n. # . .\n# # # #\nG A . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (1, 1), (1, 2), (1, 3), (2, 1), (3, 0), (3, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 3], [1, 1], [2, 0], [2, 1], [2, 2], [2, 3], [3, 3]], "position": [3, 1], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.9, "east": 0.1}, "solvable": {"false": 0.6, "true": 0.4}, "value": {"0": 0.65, "1": 0.09, "2": 0.07, "3": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:6b6c38cd29741d8e760337e1", "state_id": "grid_navigation:6b6c38cd29741d8e760337e1", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . G .\n. . . #\n# . # .\n. # . A", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.5, "west": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 0), (2, 1), (2, 3), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 3], [2, 0], [2, 2], [3, 1]], "position": [3, 3], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.25, "north": 0.75}, "solvable": {"false": 0.28, "true": 0.72}, "value": {"0": 0.01, "1": 0.34, "2": 0.54, "3": 0.11}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:47691358ae5a058c576f3481", "state_id": "tic_tac_toe:47691358ae5a058c576f3481", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X O\nX O O\n. . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place X in row 3, column 2 (cell 8).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_1": "Place X in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_7", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_8": 0.0, "cell_7": 1.0, "cell_1": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XXOO.XO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XOXOO..X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.26, "cell_1": 0.36, "cell_7": 0.38}, "solvable": {"false": 0.6699999999999999, "true": 0.33}, "value": {"0": 0.23, "1": 0.49, "2": 0.28}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:95d2e3e6ce64014051a2a1e6", "state_id": "grid_navigation:95d2e3e6ce64014051a2a1e6", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # .\nG . # #\n# . . .\n# . . A", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.5, "west": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 2), (0, 3), (2, 0), (2, 1), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [1, 2], [1, 3], [2, 0], [3, 0]], "position": [3, 3], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.49, "west": 0.51}, "solvable": {"false": 0.20999999999999996, "true": 0.79}, "value": {"0": 0.01, "1": 0.37, "2": 0.52, "3": 0.1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:1a0139b721cc41772ba7ad63", "state_id": "tic_tac_toe:1a0139b721cc41772ba7ad63", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O X\n. . X\n. X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_1": "Place O in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_7": 0.3333333333333333, "cell_5": 0.3333333333333333, "cell_4": 0.3333333333333333, "cell_1": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4", "cell_5", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...O.XXXO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".OX..X.XO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.2, "cell_1": 0.27, "cell_5": 0.33, "cell_7": 0.2}, "solvable": {"false": 0.89, "true": 0.11}, "value": {"0": 0.56, "1": 0.29, "2": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:70ae02a7f34626b7bbe0225c", "state_id": "grid_navigation:70ae02a7f34626b7bbe0225c", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n. . A .\n. # . .\n. . # G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.5, "west": 0.0, "east": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (1, 2), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [2, 1], [3, 2]], "position": [1, 2], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.07, "south": 0.31, "east": 0.62}, "solvable": {"false": 0.35, "true": 0.65}, "value": {"0": 0.02, "1": 0.52, "2": 0.32, "3": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:93e9f7ce063542de66a8083a", "state_id": "tic_tac_toe:93e9f7ce063542de66a8083a", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . X\nO . O\n. . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_8": "Place X in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.0, "cell_2": 0.5, "cell_9": 0.0, "cell_5": 0.5, "cell_8": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...O.OX.X:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.XO.O...", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.06, "cell_7": 0.15, "cell_5": 0.42, "cell_2": 0.31, "cell_8": 0.06}, "solvable": {"false": 0.56, "true": 0.44}, "value": {"0": 0.15, "1": 0.36, "2": 0.49}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:edd6c140ce43c6dbb460dc4b", "state_id": "grid_navigation:edd6c140ce43c6dbb460dc4b", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG . # #\n# . # #\n. # A .\n# # . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"south": 0.5, "east": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 3), (1, 2), (1, 3), (2, 0), (2, 1), (3, 0), (3, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [1, 0], [1, 2], [1, 3], [2, 1], [3, 0], [3, 1], [3, 3]], "position": [2, 2], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.59, "south": 0.41}, "solvable": {"false": 0.6, "true": 0.4}, "value": {"0": 0.07, "1": 0.39999999999999997, "2": 0.2, "3": 0.33}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:f42af77038450c82cf09ec76", "state_id": "tic_tac_toe:f42af77038450c82cf09ec76", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . .\nX X O\n. O X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_7": 0.3333333333333333, "cell_3": 0.3333333333333333, "cell_2": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_3", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OOXXXO.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O..XXO.OX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.26, "cell_3": 0.25, "cell_7": 0.49}, "solvable": {"false": 0.5700000000000001, "true": 0.43}, "value": {"0": 0.17, "1": 0.31, "2": 0.52}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:d9d76095c6ee5f26d2629e0a", "state_id": "tic_tac_toe:d9d76095c6ee5f26d2629e0a", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . .\nO X .\nX . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place X in row 1, column 2 (cell 2).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_3": "Place X in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.25, "cell_6": 0.0, "cell_8": 0.25, "cell_9": 0.25, "cell_3": 0.25}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_3", "cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....X.OOX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O..OX.X..", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.17, "cell_8": 0.21, "cell_3": 0.18, "cell_6": 0.22, "cell_2": 0.22}, "solvable": {"false": 0.72, "true": 0.28}, "value": {"0": 0.28, "1": 0.36, "2": 0.36}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:e2fbeaf60ba1894889921862", "state_id": "grid_navigation:e2fbeaf60ba1894889921862", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. . G .\n. . . .\n. A # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 3), (1, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[3, 2], [3, 3]], "position": [3, 1], "goal": [1, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.2, "north": 0.8}, "solvable": {"false": 0.08999999999999997, "true": 0.91}, "value": {"0": 0.06, "1": 0.74, "2": 0.17, "3": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:54e053ca14b3e03669b8de70", "state_id": "tic_tac_toe:54e053ca14b3e03669b8de70", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . .\nO X X\nX O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.0, "cell_3": 1.0, "cell_2": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXXO.OX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O..OXXXO.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.46, "cell_3": 0.32, "cell_2": 0.22}, "solvable": {"false": 0.56, "true": 0.44}, "value": {"0": 0.25, "1": 0.3, "2": 0.45}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:6f573f50c218668964f54fa8", "state_id": "grid_navigation:6f573f50c218668964f54fa8", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # .\n. # . A\n. # # #\nG # . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.5, "west": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 3), (1, 1), (1, 2), (1, 3), (2, 1), (3, 1), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [1, 1], [2, 1], [2, 2], [2, 3], [3, 1], [3, 3]], "position": [1, 3], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.3, "west": 0.7}, "solvable": {"false": 0.5700000000000001, "true": 0.43}, "value": {"0": 0.03, "1": 0.37, "2": 0.3, "3": 0.3}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:8246a605303ad706d4ee63a5", "state_id": "tic_tac_toe:8246a605303ad706d4ee63a5", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . X\nO X .\n. O X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place X in row 1, column 2 (cell 2).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_6", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.0, "cell_7": 0.5, "cell_6": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_6", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOOX.X.X:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.XOX..OX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.44, "cell_2": 0.15, "cell_7": 0.41}, "solvable": {"false": 0.5800000000000001, "true": 0.42}, "value": {"0": 0.1, "1": 0.27, "2": 0.63}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:34a7f34b10e35b4f85013f68", "state_id": "grid_navigation:34a7f34b10e35b4f85013f68", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n# # . A\n# . # .\n. . . G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "south": 1.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((1, 1), (1, 3), (2, 2), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [1, 1], [2, 0], [2, 2]], "position": [1, 3], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.71, "west": 0.2, "north": 0.09}, "solvable": {"false": 0.29000000000000004, "true": 0.71}, "value": {"0": 0.03, "1": 0.58, "2": 0.29, "3": 0.1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:7f3de6da95360cca0845c8cc", "state_id": "grid_navigation:7f3de6da95360cca0845c8cc", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. G . #\n. . . #\n. # # #\n. . A .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 2}, "gold_probs": {"action": {"west": 1.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 3), (2, 1), (2, 2), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 3], [2, 1], [2, 2], [2, 3]], "position": [3, 2], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.68, "east": 0.32}, "solvable": {"false": 0.39, "true": 0.61}, "value": {"0": 0.02, "1": 0.4, "2": 0.4, "3": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:b728a2584b51463acfa92f75", "state_id": "grid_navigation:b728a2584b51463acfa92f75", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # . .\n. # . #\n# G A .\n. . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east.", "west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "east": 0.0, "west": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (0, 3), (1, 2), (1, 3), (3, 0), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [1, 1], [1, 3], [2, 0], [3, 3]], "position": [2, 2], "goal": [2, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.96, "north": 0.03, "east": 0, "south": 0.01}, "solvable": {"false": 0.14, "true": 0.86}, "value": {"0": 0.8099999999999999, "1": 0.12, "2": 0.03, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:e997510c457f4afa570484d1", "state_id": "tic_tac_toe:e997510c457f4afa570484d1", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X X\nO O .\nX O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_6": "Place O in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_6", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.0, "cell_6": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXXOOX.X:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XXOO.XOX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.44, "cell_6": 0.56}, "solvable": {"false": 0.8200000000000001, "true": 0.18}, "value": {"0": 0.4, "1": 0.36, "2": 0.24}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:fb0f0bb674fdafb2a8b06338", "state_id": "tic_tac_toe:fb0f0bb674fdafb2a8b06338", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X .\nO . X\nO X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place O in row 1, column 3 (cell 3).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_3": 0.5, "cell_5": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.XOX.XXOO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XX.O.XOXO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.5, "cell_3": 0.5}, "solvable": {"false": 0.86, "true": 0.14}, "value": {"0": 0.33, "1": 0.48, "2": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:a6434e9b3ef1a44a963067c2", "state_id": "tic_tac_toe:a6434e9b3ef1a44a963067c2", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . .\n. X O\nX . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_3": 0.25, "cell_8": 0.25, "cell_2": 0.25, "cell_9": 0.25}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_3", "cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OOX...X:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O...XOX..", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.2, "cell_8": 0.19, "cell_9": 0.18, "cell_4": 0.3, "cell_3": 0.13}, "solvable": {"false": 0.7, "true": 0.3}, "value": {"0": 0.16, "1": 0.42, "2": 0.42}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:2de457280be0155f5f06ffa7", "state_id": "grid_navigation:2de457280be0155f5f06ffa7", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A # .\n. . . .\n. . . .\n# . . G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 0.0, "south": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 3), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [3, 0]], "position": [0, 1], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.07, "south": 0.93}, "solvable": {"false": 0.09999999999999998, "true": 0.9}, "value": {"0": 0, "1": 0.3, "2": 0.67, "3": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:1922f30124df91bc340ea449", "state_id": "tic_tac_toe:1922f30124df91bc340ea449", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X O\n. O X\nX O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_4": "Place O in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_1": 0.5, "cell_4": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_4"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XXOOOXX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XO.OXXOX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.62, "cell_4": 0.38}, "solvable": {"false": 0.83, "true": 0.17}, "value": {"0": 0.3, "1": 0.46, "2": 0.24}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:f6feed89314c57bea897dc48", "state_id": "tic_tac_toe:f6feed89314c57bea897dc48", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . O\nO X .\nX X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place O in row 2, column 3 (cell 6).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_6", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 1.0, "cell_2": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.O.XXXOX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.OOX.XXO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.43, "cell_6": 0.57}, "solvable": {"false": 0.89, "true": 0.11}, "value": {"0": 0.56, "1": 0.32, "2": 0.11}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:c5c88053e710617c91421a1c", "state_id": "tic_tac_toe:c5c88053e710617c91421a1c", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O O\nX . X\n. O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.5, "cell_5": 0.5, "cell_9": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.O.X.XOOX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOOX.X.O.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.25, "cell_7": 0.27, "cell_5": 0.48}, "solvable": {"false": 0.45999999999999996, "true": 0.54}, "value": {"0": 0.16, "1": 0.3, "2": 0.54}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:6de8540c73b19279e966e3c6", "state_id": "tic_tac_toe:6de8540c73b19279e966e3c6", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . X\n. O O\nX . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place O in row 1, column 2 (cell 2).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_8": "Place O in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.0, "cell_4": 1.0, "cell_9": 0.0, "cell_8": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOO.X.X:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.X.OOX..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.32, "cell_8": 0.26, "cell_9": 0.14, "cell_4": 0.28}, "solvable": {"false": 0.8200000000000001, "true": 0.18}, "value": {"0": 0.49, "1": 0.27, "2": 0.24}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:5222beca8ab62ddf188cd814", "state_id": "tic_tac_toe:5222beca8ab62ddf188cd814", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X O\n. . O\nO . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_8": 0.0, "cell_5": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.X..OXXO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XXO..OO.X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.46, "cell_8": 0.32, "cell_4": 0.22}, "solvable": {"false": 0.74, "true": 0.26}, "value": {"0": 0.32, "1": 0.46, "2": 0.22}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:77c77e710749bcc5d15791c5", "state_id": "tic_tac_toe:77c77e710749bcc5d15791c5", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O .\nO X X\nX O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_1": 0.5, "cell_3": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.O.OXXXOX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".O.OXXXOX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.67, "cell_3": 0.33}, "solvable": {"false": 0.85, "true": 0.15}, "value": {"0": 0.45, "1": 0.39, "2": 0.16}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:c85248c52c8205c9082c9cfb", "state_id": "grid_navigation:c85248c52c8205c9082c9cfb", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n# # . A\n. # # .\n. . G #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.3333333333333333, "west": 0.3333333333333333, "south": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (1, 1), (1, 2), (2, 2), (2, 3), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 0], [1, 1], [2, 1], [2, 2], [3, 3]], "position": [1, 3], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.12, "west": 0.31, "south": 0.57}, "solvable": {"false": 0.51, "true": 0.49}, "value": {"0": 0.02, "1": 0.43, "2": 0.36, "3": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:5f9d567edeba14850bba8e34", "state_id": "grid_navigation:5f9d567edeba14850bba8e34", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\nA . # .\n# # G .\n. # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 2}, "gold_probs": {"action": {"north": 0.5, "east": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 2), (1, 2), (1, 3), (2, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 2], [2, 0], [2, 1], [3, 1]], "position": [1, 0], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.79, "north": 0.21}, "solvable": {"false": 0.5, "true": 0.5}, "value": {"0": 0.08, "1": 0.54, "2": 0.2, "3": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:e6b38aecfa46e3e9235fc069", "state_id": "grid_navigation:e6b38aecfa46e3e9235fc069", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # G\n. . . .\n. # . .\n# # . A", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 1.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 3), (2, 2), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [2, 1], [3, 0], [3, 1]], "position": [3, 3], "goal": [0, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.77, "west": 0.23}, "solvable": {"false": 0.22999999999999998, "true": 0.77}, "value": {"0": 0.01, "1": 0.33, "2": 0.56, "3": 0.1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:64c523c1d7db7d52a899c2b0", "state_id": "grid_navigation:64c523c1d7db7d52a899c2b0", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# A . #\n. . . G\n. # . #\n# # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.5, "south": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 2), (1, 3), (2, 2), (2, 3), (3, 0), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 3], [2, 1], [2, 3], [3, 0], [3, 1], [3, 2]], "position": [0, 1], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.32, "south": 0.68}, "solvable": {"false": 0.09999999999999998, "true": 0.9}, "value": {"0": 0.11, "1": 0.79, "2": 0.08, "3": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:8703eff38e72b6bfab232711", "state_id": "tic_tac_toe:8703eff38e72b6bfab232711", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . O\n. X .\nO X X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_6": 0.0, "cell_2": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.O.X.OXX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.O.X.OXX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.3, "cell_4": 0.41, "cell_2": 0.29}, "solvable": {"false": 0.54, "true": 0.46}, "value": {"0": 0.34, "1": 0.27, "2": 0.39}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:9ffe3ecc3d3076af80fab07d", "state_id": "grid_navigation:9ffe3ecc3d3076af80fab07d", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A . .\n. . . .\n# G # #\n# . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "east": 0.0, "south": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 1), (2, 1), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[2, 0], [2, 2], [2, 3], [3, 0]], "position": [0, 1], "goal": [2, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.04, "west": 0.28, "south": 0.68}, "solvable": {"false": 0.26, "true": 0.74}, "value": {"0": 0.06, "1": 0.51, "2": 0.32, "3": 0.11}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:822d1dc651a3b0407de84803", "state_id": "tic_tac_toe:822d1dc651a3b0407de84803", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X O\n. . .\nX O X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_4": 0.3333333333333333, "cell_6": 0.3333333333333333, "cell_5": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4", "cell_5", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.XX.OO.X:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXO...XOX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.41, "cell_6": 0.16, "cell_4": 0.43}, "solvable": {"false": 0.75, "true": 0.25}, "value": {"0": 0.23, "1": 0.53, "2": 0.24}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:1ffaedb7720546d8ff5be502", "state_id": "tic_tac_toe:1ffaedb7720546d8ff5be502", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . X\n. O O\nX X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_1": 0.3333333333333333, "cell_4": 0.3333333333333333, "cell_2": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_2", "cell_4"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..X.OOXXO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..X.OOXXO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.42, "cell_2": 0.25, "cell_1": 0.33}, "solvable": {"false": 0.5800000000000001, "true": 0.42}, "value": {"0": 0.13, "1": 0.34, "2": 0.53}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:9c82c91fec3f5373202069e1", "state_id": "tic_tac_toe:9c82c91fec3f5373202069e1", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . O\nO X X\n. . X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place O in row 3, column 2 (cell 8).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_8": 0.0, "cell_1": 1.0, "cell_7": 0.0, "cell_2": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OOXX..X:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..OOXX..X", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.27, "cell_7": 0.28, "cell_8": 0.22, "cell_2": 0.23}, "solvable": {"false": 0.86, "true": 0.14}, "value": {"0": 0.54, "1": 0.28, "2": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:a7695cf6aad9741e51d8bd0a", "state_id": "tic_tac_toe:a7695cf6aad9741e51d8bd0a", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\nX O X\nX . O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_8": "Place O in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 1.0, "cell_3": 0.0, "cell_2": 0.0, "cell_8": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...XOXO.X:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "...XOXX.O", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.1, "cell_8": 0.51, "cell_2": 0.2, "cell_1": 0.19}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.45, "1": 0.38, "2": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:b6b2ad8a809971a221cab9fd", "state_id": "tic_tac_toe:b6b2ad8a809971a221cab9fd", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . O\nO . .\nX . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place X in row 1, column 2 (cell 2).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.0, "cell_5": 0.0, "cell_8": 0.0, "cell_6": 0.0, "cell_9": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O...XOX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.OO..X..", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.07, "cell_8": 0.11, "cell_9": 0.1, "cell_2": 0.17, "cell_5": 0.55}, "solvable": {"false": 0.73, "true": 0.27}, "value": {"0": 0.18, "1": 0.44, "2": 0.38}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:e4c86c8296f3846c91a2ea7a", "state_id": "tic_tac_toe:e4c86c8296f3846c91a2ea7a", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . X\n. O O\nX . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_8", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_8": 1.0, "cell_2": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.X.O.XOX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.X.OOX.X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.37, "cell_8": 0.42, "cell_2": 0.21}, "solvable": {"false": 0.51, "true": 0.49}, "value": {"0": 0.15, "1": 0.28, "2": 0.57}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:90495ff76523f2b3ae3484c6", "state_id": "grid_navigation:90495ff76523f2b3ae3484c6", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # G\n. # . #\n. . . #\nA . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 2), (0, 3), (1, 0), (2, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 1], [1, 3], [2, 3], [3, 3]], "position": [3, 0], "goal": [0, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.49, "east": 0.51}, "solvable": {"false": 0.42000000000000004, "true": 0.58}, "value": {"0": 0.01, "1": 0.43, "2": 0.32, "3": 0.24}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:c51b70978c446ab07544b98c", "state_id": "tic_tac_toe:c51b70978c446ab07544b98c", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . X\n. . O\n. X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place X in row 2, column 2 (cell 5).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 0.5, "cell_9": 0.0, "cell_4": 0.0, "cell_2": 0.0, "cell_7": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OX...OX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.X..O.X.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.12, "cell_2": 0.1, "cell_7": 0.16, "cell_5": 0.46, "cell_9": 0.16}, "solvable": {"false": 0.7, "true": 0.3}, "value": {"0": 0.15, "1": 0.41, "2": 0.44}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:f0dc74415f5b09a90d3f30d9", "state_id": "grid_navigation:f0dc74415f5b09a90d3f30d9", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # A G\n. . . .\n. # . .\n. # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (0, 3), (2, 2), (3, 1), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [2, 1], [3, 1], [3, 2]], "position": [0, 2], "goal": [0, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.93, "south": 0.07}, "solvable": {"false": 0.12, "true": 0.88}, "value": {"0": 0.95, "1": 0.02, "2": 0.01, "3": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:23e93dded29761fb1cd0d702", "state_id": "grid_navigation:23e93dded29761fb1cd0d702", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # .\nG . # .\n# # # A\n. . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"south": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 0), (1, 2), (2, 0), (2, 1), (2, 2), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [1, 2], [2, 0], [2, 1], [2, 2], [3, 2]], "position": [2, 3], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.58, "south": 0.42}, "solvable": {"false": 0.62, "true": 0.38}, "value": {"0": 0.05, "1": 0.29, "2": 0.25, "3": 0.41}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:912d5f77fd2f912a242e5d46", "state_id": "grid_navigation:912d5f77fd2f912a242e5d46", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. G . #\n. # A .\n# . . .\n. . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.0, "south": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 1), (2, 0), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 1], [2, 0], [3, 2], [3, 3]], "position": [1, 2], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.67, "south": 0.15, "east": 0.18}, "solvable": {"false": 0.47, "true": 0.53}, "value": {"0": 0.23, "1": 0.51, "2": 0.15, "3": 0.11}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:8ffce3601b163c2a1865b08f", "state_id": "grid_navigation:8ffce3601b163c2a1865b08f", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# G # .\n. . A .\n# # . .\n. . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 1.0, "south": 0.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 2), (2, 0), (2, 1), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [2, 0], [2, 1], [3, 2]], "position": [1, 2], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.1, "west": 0.84, "south": 0.06}, "solvable": {"false": 0.4, "true": 0.6}, "value": {"0": 0.14, "1": 0.55, "2": 0.18, "3": 0.13}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:58c191be3bddd73a7339742d", "state_id": "grid_navigation:58c191be3bddd73a7339742d", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG . A #\n# . . #\n# . # .\n. # . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 2), (1, 3), (2, 2), (3, 0), (3, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 0], [1, 3], [2, 0], [2, 2], [3, 1], [3, 3]], "position": [0, 2], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.9, "south": 0.1}, "solvable": {"false": 0.28, "true": 0.72}, "value": {"0": 0.64, "1": 0.23, "2": 0.07, "3": 0.06}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:6024541b6200c0d6ebffbb63", "state_id": "tic_tac_toe:6024541b6200c0d6ebffbb63", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . O\n. X X\nO X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_4": "Place X in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.0, "cell_2": 0.5, "cell_4": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_4"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.XOXX.O.O:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.O.XXOX.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.33, "cell_2": 0.22, "cell_9": 0.45}, "solvable": {"false": 0.55, "true": 0.45}, "value": {"0": 0.35, "1": 0.29, "2": 0.36}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:70360cac40f430a4b446c160", "state_id": "tic_tac_toe:70360cac40f430a4b446c160", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X .\nO O X\nX . X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place O in row 3, column 2 (cell 8).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_8": 0.5, "cell_3": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.XOXOOX.X:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OX.OOXX.X", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.38, "cell_8": 0.62}, "solvable": {"false": 0.81, "true": 0.19}, "value": {"0": 0.39999999999999997, "1": 0.38, "2": 0.22}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:346443c941d64b4e1b412f5b", "state_id": "tic_tac_toe:346443c941d64b4e1b412f5b", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O X\nO . O\n. X X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.5, "cell_1": 0.0, "cell_5": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.O.O.XXOX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".OXO.O.XX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.32, "cell_1": 0.23, "cell_5": 0.45}, "solvable": {"false": 0.5700000000000001, "true": 0.43}, "value": {"0": 0.27, "1": 0.33, "2": 0.4}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:4c7db53b1fe0854ae5e32c9d", "state_id": "tic_tac_toe:4c7db53b1fe0854ae5e32c9d", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O O\nX . X\nO . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.0, "cell_8": 0.0, "cell_5": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OX.XOOX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOOX.XO..", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.2, "cell_8": 0.27, "cell_5": 0.53}, "solvable": {"false": 0.44999999999999996, "true": 0.55}, "value": {"0": 0.13, "1": 0.24, "2": 0.63}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:a8476ee0b29d8d3891800dd6", "state_id": "grid_navigation:a8476ee0b29d8d3891800dd6", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # G A\n. . . .\n. # . #\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (2, 0), (2, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [2, 1], [2, 3]], "position": [0, 3], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.1, "west": 0.9}, "solvable": {"false": 0.45999999999999996, "true": 0.54}, "value": {"0": 0.75, "1": 0.14, "2": 0.05, "3": 0.06}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:f4e2b1057c846df2bafbec3f", "state_id": "grid_navigation:f4e2b1057c846df2bafbec3f", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n# . . #\nG # # .\n. A . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 1.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 1), (2, 1), (2, 3), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 0], [1, 3], [2, 1], [2, 2]], "position": [3, 1], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.54, "west": 0.46}, "solvable": {"false": 0.64, "true": 0.36}, "value": {"0": 0.02, "1": 0.3, "2": 0.4, "3": 0.28}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:33c67a10380ab0ca1905acca", "state_id": "tic_tac_toe:33c67a10380ab0ca1905acca", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO O .\nX X O\nX . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place X in row 3, column 2 (cell 8).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_3": "Place X in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_8": 0.0, "cell_9": 0.0, "cell_3": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOXX.OO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OO.XXOX..", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.58, "cell_3": 0.22, "cell_9": 0.2}, "solvable": {"false": 0.41000000000000003, "true": 0.59}, "value": {"0": 0.12, "1": 0.15, "2": 0.73}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:18c2e12834f222450eebf8df", "state_id": "grid_navigation:18c2e12834f222450eebf8df", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . A\n# # G .\n# . # .\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.5, "west": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 2), (1, 2), (1, 3), (2, 1), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 0], [1, 1], [2, 0], [2, 2]], "position": [0, 3], "goal": [1, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.44, "south": 0.56}, "solvable": {"false": 0.51, "true": 0.49}, "value": {"0": 0.22, "1": 0.37, "2": 0.2, "3": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:c7a9ecf22ac5307e3d153fda", "state_id": "tic_tac_toe:c7a9ecf22ac5307e3d153fda", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X X\nO . O\nX X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.0, "cell_5": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXX.XXOO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XXO.OXXO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.3, "cell_5": 0.7}, "solvable": {"false": 0.83, "true": 0.17}, "value": {"0": 0.49, "1": 0.31, "2": 0.2}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:c1aaecf0c50d6696f6d55238", "state_id": "grid_navigation:c1aaecf0c50d6696f6d55238", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . G .\n# # # A\n# . . .\n# # # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 1), (1, 2), (1, 3), (2, 3), (3, 0), (3, 1), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [1, 1], [1, 2], [2, 0], [3, 0], [3, 1], [3, 2], [3, 3]], "position": [1, 3], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.2, "north": 0.8}, "solvable": {"false": 0.43000000000000005, "true": 0.57}, "value": {"0": 0.36, "1": 0.4, "2": 0.1, "3": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:52d90a27ad45811cd4078ce9", "state_id": "tic_tac_toe:52d90a27ad45811cd4078ce9", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . O\nX X .\n. X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place O in row 2, column 3 (cell 6).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.0, "cell_7": 0.0, "cell_2": 1.0, "cell_9": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXX..XO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.OXX..X.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.21, "cell_2": 0.1, "cell_7": 0.19, "cell_6": 0.5}, "solvable": {"false": 0.81, "true": 0.19}, "value": {"0": 0.36, "1": 0.36, "2": 0.28}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:349f7815bf30e646b898e5e9", "state_id": "grid_navigation:349f7815bf30e646b898e5e9", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # .\n. . G .\n. . . A\n# # . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.5, "north": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (0, 2), (3, 0), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [3, 0], [3, 1], [3, 3]], "position": [2, 3], "goal": [1, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.64, "west": 0.36}, "solvable": {"false": 0.07999999999999996, "true": 0.92}, "value": {"0": 0.39, "1": 0.5599999999999999, "2": 0.04, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:a50bfff5806b6d947624fc3c", "state_id": "tic_tac_toe:a50bfff5806b6d947624fc3c", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . X\nO X .\nO . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_6": "Place O in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_9": 0.25, "cell_2": 0.25, "cell_8": 0.25, "cell_6": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_6", "cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O.XOX.X:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.XOX.O..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.18, "cell_2": 0.32, "cell_6": 0.31, "cell_8": 0.19}, "solvable": {"false": 0.86, "true": 0.14}, "value": {"0": 0.51, "1": 0.3, "2": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:8e8e1181b22a55a65495e2a9", "state_id": "grid_navigation:8e8e1181b22a55a65495e2a9", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A . .\n. G . .\n. . # .\n. # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 1.0, "west": 0.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((1, 3), (2, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[2, 2], [3, 1]], "position": [0, 1], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.01, "west": 0.17, "south": 0.82}, "solvable": {"false": 0.06999999999999995, "true": 0.93}, "value": {"0": 0.99, "1": 0.01, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:e744909b58ff262be2fb324d", "state_id": "tic_tac_toe:e744909b58ff262be2fb324d", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X .\nO X .\nX . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_9": 0.25, "cell_6": 0.25, "cell_8": 0.25, "cell_3": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_6", "cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....XXXOO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OX.OX.X..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.18, "cell_8": 0.38, "cell_3": 0.21, "cell_6": 0.23}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.46, "1": 0.35, "2": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:96b8b24c6f6b6597349ea9ec", "state_id": "grid_navigation:96b8b24c6f6b6597349ea9ec", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # #\n. . # #\n. A . G\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west.", "south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "west": 0.0, "south": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (0, 3), (1, 2), (1, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [1, 2], [1, 3]], "position": [2, 1], "goal": [2, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.01, "west": 0, "east": 0.93, "south": 0.05}, "solvable": {"false": 0.13, "true": 0.87}, "value": {"0": 0.86, "1": 0.09, "2": 0.02, "3": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:e465b983eb6e933255f8add7", "state_id": "tic_tac_toe:e465b983eb6e933255f8add7", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . .\nX . .\nO X X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place O in row 1, column 2 (cell 2).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.0, "cell_6": 0.0, "cell_3": 1.0, "cell_5": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O..XXXO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O..X..OXX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.41, "cell_5": 0.37, "cell_3": 0.14, "cell_2": 0.08}, "solvable": {"false": 0.83, "true": 0.17}, "value": {"0": 0.51, "1": 0.22, "2": 0.27}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:cd99ed7515b0ac7ab108aac4", "state_id": "grid_navigation:cd99ed7515b0ac7ab108aac4", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\n. # G .\n. . # #\nA . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 2), (0, 3), (1, 2), (2, 0), (2, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 1], [2, 2], [2, 3], [3, 3]], "position": [3, 0], "goal": [1, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.51, "east": 0.49}, "solvable": {"false": 0.29000000000000004, "true": 0.71}, "value": {"0": 0.01, "1": 0.54, "2": 0.29, "3": 0.16}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:21e7ec23fbe847eb86ec040b", "state_id": "grid_navigation:21e7ec23fbe847eb86ec040b", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. . # .\n. . A .\nG . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 0.5, "east": 0.0, "south": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((2, 2),)", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 2]], "position": [2, 2], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.01, "south": 0.49, "west": 0.5}, "solvable": {"false": 0.13, "true": 0.87}, "value": {"0": 0.09, "1": 0.73, "2": 0.15, "3": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:4067b2abe75d778c1f80e149", "state_id": "grid_navigation:4067b2abe75d778c1f80e149", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # #\n# # A .\n. # G #\n. . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.0, "south": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (0, 3), (1, 0), (1, 3), (2, 1), (2, 2), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [1, 0], [1, 1], [2, 1], [2, 3], [3, 2]], "position": [1, 2], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.56, "east": 0.44}, "solvable": {"false": 0.5700000000000001, "true": 0.43}, "value": {"0": 0.27, "1": 0.32, "2": 0.12, "3": 0.29}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:be6a4e2e65bc8a51a009781e", "state_id": "tic_tac_toe:be6a4e2e65bc8a51a009781e", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . .\nX . X\nO . O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place O in row 2, column 2 (cell 5).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_8": "Place O in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 0.5, "cell_3": 0.0, "cell_2": 0.0, "cell_8": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XX.XO.O:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X..X.XO.O", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.14, "cell_8": 0.17, "cell_2": 0.29, "cell_5": 0.4}, "solvable": {"false": 0.79, "true": 0.21}, "value": {"0": 0.41, "1": 0.3, "2": 0.29}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:dd957cde31463ae8c47ec3fc", "state_id": "grid_navigation:dd957cde31463ae8c47ec3fc", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . A\nG # . .\n. # . .\n. # # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": false, "value": 3}, "gold_probs": {"action": {"south": 0.5, "west": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (1, 1), (1, 2), (1, 3), (2, 3), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 1], [2, 1], [3, 1], [3, 2], [3, 3]], "position": [0, 3], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.59, "west": 0.41}, "solvable": {"false": 0.56, "true": 0.44}, "value": {"0": 0.03, "1": 0.32, "2": 0.4, "3": 0.25}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:0f2af7a58b545c7f68461f49", "state_id": "grid_navigation:0f2af7a58b545c7f68461f49", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG # . #\n# # A #\n. # . #\n. # . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"south": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 3), (1, 0), (1, 1), (1, 3), (2, 1), (2, 3), (3, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 3], [1, 0], [1, 1], [1, 3], [2, 1], [2, 3], [3, 1], [3, 3]], "position": [1, 2], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.21, "north": 0.79}, "solvable": {"false": 0.79, "true": 0.21}, "value": {"0": 0.19, "1": 0.1, "2": 0.06, "3": 0.65}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:e820fad2f66b0ef1d1beb01f", "state_id": "tic_tac_toe:e820fad2f66b0ef1d1beb01f", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O .\n. X .\nX O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place O in row 1, column 3 (cell 3).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_4": "Place O in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_3": 0.25, "cell_1": 0.25, "cell_6": 0.25, "cell_4": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_3", "cell_4", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOXO..X:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".O..X.XOX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.29, "cell_3": 0.16, "cell_1": 0.35, "cell_6": 0.2}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.41, "1": 0.43, "2": 0.16}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:5023e95cc7b84203dcc2d5d5", "state_id": "tic_tac_toe:5023e95cc7b84203dcc2d5d5", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . O\n. . X\n. . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_4": 0.2, "cell_2": 0.2, "cell_7": 0.2, "cell_8": 0.2, "cell_5": 0.2}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_4", "cell_5", "cell_7", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O...XXO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.O..X..X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.14, "cell_2": 0.15, "cell_7": 0.22, "cell_4": 0.15, "cell_5": 0.33999999999999997}, "solvable": {"false": 0.54, "true": 0.46}, "value": {"0": 0.12, "1": 0.24, "2": 0.64}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:557f4e1a12fb5d41248b9814", "state_id": "tic_tac_toe:557f4e1a12fb5d41248b9814", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X .\n. O X\n. X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place O in row 2, column 1 (cell 4).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_1": "Place O in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_7": 0.3333333333333333, "cell_3": 0.3333333333333333, "cell_1": 0.3333333333333333}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_3", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...XOX.XO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".X..OX.XO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.27, "cell_4": 0.3, "cell_3": 0.17, "cell_7": 0.26}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.42, "1": 0.37, "2": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:fa8ba402b229989fe8e4254c", "state_id": "grid_navigation:fa8ba402b229989fe8e4254c", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . .\n. # # .\n. # # #\nA . . G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (1, 1), (1, 2), (2, 1), (2, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 1], [1, 2], [2, 1], [2, 2], [2, 3]], "position": [3, 0], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.11, "east": 0.89}, "solvable": {"false": 0.29000000000000004, "true": 0.71}, "value": {"0": 0.09, "1": 0.59, "2": 0.21, "3": 0.11}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:0571aeb59347aa0d61e5a8aa", "state_id": "grid_navigation:0571aeb59347aa0d61e5a8aa", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . #\nA . . .\nG # . #\n# . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.0, "south": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 3), (1, 1), (3, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 3], [2, 1], [2, 3], [3, 0]], "position": [1, 0], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.15, "south": 0.85}, "solvable": {"false": 0.31999999999999995, "true": 0.68}, "value": {"0": 0.26, "1": 0.5, "2": 0.17, "3": 0.07}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:853087769d297de6d2b421e1", "state_id": "grid_navigation:853087769d297de6d2b421e1", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\n. . # G\n. . A #\n. # . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": false, "value": 3}, "gold_probs": {"action": {"south": 0.5, "west": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (0, 3), (1, 1), (2, 0), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 2], [2, 3], [3, 1], [3, 3]], "position": [2, 2], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.13, "west": 0.87}, "solvable": {"false": 0.33999999999999997, "true": 0.66}, "value": {"0": 0.19, "1": 0.51, "2": 0.14, "3": 0.16}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:67895f76529d9ecf960fe4a7", "state_id": "grid_navigation:67895f76529d9ecf960fe4a7", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . A .\n# # # #\n. . . G\n. . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "west": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 2), (1, 2), (2, 2), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 0], [1, 1], [1, 2], [1, 3], [3, 3]], "position": [0, 2], "goal": [2, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.26, "east": 0.74}, "solvable": {"false": 0.55, "true": 0.45}, "value": {"0": 0.04, "1": 0.31, "2": 0.27, "3": 0.38}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:a7fd9ac09afb1cd12ff59bfa", "state_id": "tic_tac_toe:a7fd9ac09afb1cd12ff59bfa", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O X\n. O X\nX . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place O in row 2, column 1 (cell 4).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_1": "Place O in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_8", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_8": 0.5, "cell_9": 0.5, "cell_1": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOO.XX.:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".OX.OXX..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.22, "cell_8": 0.37, "cell_9": 0.17, "cell_1": 0.24}, "solvable": {"false": 0.84, "true": 0.16}, "value": {"0": 0.52, "1": 0.26, "2": 0.22}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:8ac484b30aa30bd8a5ca082b", "state_id": "tic_tac_toe:8ac484b30aa30bd8a5ca082b", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X O\n. . X\nX O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place O in row 2, column 1 (cell 4).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_4": 0.5, "cell_5": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4", "cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.XX.OOXX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXO..XXOX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.47, "cell_5": 0.53}, "solvable": {"false": 0.91, "true": 0.09}, "value": {"0": 0.34, "1": 0.48, "2": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:7e1a227cd05e0ddac70f67b3", "state_id": "tic_tac_toe:7e1a227cd05e0ddac70f67b3", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X X\n. O X\n. O O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_4": "Place X in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 1.0, "cell_7": 0.0, "cell_4": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...OOXOXX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XX.OX.OO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.23, "cell_7": 0.35, "cell_1": 0.42}, "solvable": {"false": 0.45999999999999996, "true": 0.54}, "value": {"0": 0.21, "1": 0.28, "2": 0.51}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:e3fd4d85da3408fed7edb229", "state_id": "grid_navigation:e3fd4d85da3408fed7edb229", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . G .\n. . . #\n. . . .\n. . . A", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 0.5, "north": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0),)", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 3]], "position": [3, 3], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.76, "west": 0.24}, "solvable": {"false": 0.08999999999999997, "true": 0.91}, "value": {"0": 0, "1": 0.47, "2": 0.48, "3": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:e98e9790d3b68daf33a556ca", "state_id": "grid_navigation:e98e9790d3b68daf33a556ca", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. . . #\n. A . .\n. . . G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west.", "north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.5, "west": 0.0, "north": 0.0, "east": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2),)", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 3]], "position": [2, 1], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.38, "west": 0.01, "north": 0.02, "east": 0.59}, "solvable": {"false": 0.07999999999999996, "true": 0.92}, "value": {"0": 0.14, "1": 0.68, "2": 0.13, "3": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:a85d6de741ebd6a87e645fc8", "state_id": "grid_navigation:a85d6de741ebd6a87e645fc8", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . A .\n. # . .\n. . # .\n. . G #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.3333333333333333, "west": 0.3333333333333333, "south": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (1, 1), (2, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 1], [2, 2], [3, 3]], "position": [0, 2], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.37, "south": 0.26, "west": 0.37}, "solvable": {"false": 0.29000000000000004, "true": 0.71}, "value": {"0": 0.01, "1": 0.53, "2": 0.35, "3": 0.11}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:4379c1da53580baa1ea9ffc0", "state_id": "tic_tac_toe:4379c1da53580baa1ea9ffc0", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . O\nO X X\n. X X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place O in row 1, column 2 (cell 2).", "cell_7": "Place O in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.5, "cell_7": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOXX.XXO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.OOXX.XX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.75, "cell_2": 0.25}, "solvable": {"false": 0.72, "true": 0.28}, "value": {"0": 0.36, "1": 0.25, "2": 0.39}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:453f7253a4145447a804ac50", "state_id": "tic_tac_toe:453f7253a4145447a804ac50", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . .\nO O .\nX X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place X in row 1, column 3 (cell 3).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_6", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_3": 0.0, "cell_2": 0.0, "cell_6": 1.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O.OXXOX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X..OO.XXO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.49, "cell_3": 0.34, "cell_2": 0.17}, "solvable": {"false": 0.51, "true": 0.49}, "value": {"0": 0.16, "1": 0.25, "2": 0.59}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:f79276f2329367dda034780a", "state_id": "tic_tac_toe:f79276f2329367dda034780a", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X O\n. O X\nX X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place O in row 2, column 1 (cell 4).", "cell_1": "Place O in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_1": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XXOXOXO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XO.OXXXO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.37, "cell_1": 0.63}, "solvable": {"false": 0.84, "true": 0.16}, "value": {"0": 0.46, "1": 0.33, "2": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:cebf51138d76ac2ca2a2349f", "state_id": "tic_tac_toe:cebf51138d76ac2ca2a2349f", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O X\nO X O\n. X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.3333333333333333, "cell_1": 0.3333333333333333, "cell_9": 0.3333333333333333}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_7", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.O.OXXXO.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".OXOXO.X.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.35, "cell_9": 0.3, "cell_7": 0.35}, "solvable": {"false": 0.63, "true": 0.37}, "value": {"0": 0.14, "1": 0.37, "2": 0.49}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:99395f28591231c9f79ced6b", "state_id": "grid_navigation:99395f28591231c9f79ced6b", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . #\n# . G A\n# . . .\n. . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 3), (1, 3), (2, 3), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 3], [1, 0], [2, 0], [3, 2]], "position": [1, 3], "goal": [1, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.01, "west": 0.99}, "solvable": {"false": 0.06999999999999995, "true": 0.93}, "value": {"0": 0.98, "1": 0.01, "2": 0, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:a07672706e3e4dfd24ea398c", "state_id": "tic_tac_toe:a07672706e3e4dfd24ea398c", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X O\nX . .\nO X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place O in row 2, column 2 (cell 5).", "cell_6": "Place O in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 0.5, "cell_6": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.OX.XOXX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XXOX..OXO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.27, "cell_5": 0.73}, "solvable": {"false": 0.85, "true": 0.15}, "value": {"0": 0.43, "1": 0.39, "2": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:6281f5fb57339ec58403f817", "state_id": "grid_navigation:6281f5fb57339ec58403f817", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n. . . #\n. . # #\nA . G .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (1, 1), (2, 0), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 3], [2, 2], [2, 3]], "position": [3, 0], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.24, "east": 0.76}, "solvable": {"false": 0.18999999999999995, "true": 0.81}, "value": {"0": 0.51, "1": 0.3, "2": 0.14, "3": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:ae050018414b995633fc897e", "state_id": "tic_tac_toe:ae050018414b995633fc897e", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X O\n. . .\nX X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place O in row 2, column 3 (cell 6).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_6": 0.25, "cell_4": 0.25, "cell_9": 0.25, "cell_5": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4", "cell_5", "cell_6", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OX.XX.O:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXO...XX.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.18, "cell_6": 0.21, "cell_9": 0.54, "cell_4": 0.07}, "solvable": {"false": 0.8, "true": 0.2}, "value": {"0": 0.39, "1": 0.38, "2": 0.23}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:e9335ec402f20d2a62261d85", "state_id": "grid_navigation:e9335ec402f20d2a62261d85", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . #\n. . # .\n. A . #\n. . . G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south.", "west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.5, "south": 0.5, "west": 0.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 3), (1, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 2], [2, 3]], "position": [2, 1], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.02, "east": 0.3, "west": 0.03, "south": 0.65}, "solvable": {"false": 0.19999999999999996, "true": 0.8}, "value": {"0": 0.09, "1": 0.65, "2": 0.17, "3": 0.09}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:8b8351b5d3db0038fc5f0a9f", "state_id": "grid_navigation:8b8351b5d3db0038fc5f0a9f", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n# G # #\n. . # .\n. . . A", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (2, 1), (2, 2), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [1, 2], [1, 3], [2, 2]], "position": [3, 3], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.48, "north": 0.52}, "solvable": {"false": 0.44999999999999996, "true": 0.55}, "value": {"0": 0.01, "1": 0.34, "2": 0.46, "3": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:a003f6e31e40024b96bff3d0", "state_id": "tic_tac_toe:a003f6e31e40024b96bff3d0", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . O\nX O X\nX O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 1.0, "cell_9": 0.0, "cell_2": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXOXXO.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..OXOXXO.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.24, "cell_2": 0.1, "cell_9": 0.66}, "solvable": {"false": 0.6599999999999999, "true": 0.34}, "value": {"0": 0.34, "1": 0.42, "2": 0.24}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:60788a4492f80a1b93589d4a", "state_id": "tic_tac_toe:60788a4492f80a1b93589d4a", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . X\nO . X\nX X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place O in row 1, column 2 (cell 2).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.0, "cell_5": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.XO.XXXO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.XO.XXXO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.56, "cell_2": 0.44}, "solvable": {"false": 0.83, "true": 0.17}, "value": {"0": 0.48, "1": 0.3, "2": 0.22}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:a879372760120a38a81aa332", "state_id": "grid_navigation:a879372760120a38a81aa332", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # . .\n# . A #\n. # # .\n# . G #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 3), (1, 1), (1, 2), (2, 0), (2, 3), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [1, 0], [1, 3], [2, 1], [2, 2], [3, 0], [3, 3]], "position": [1, 2], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.53, "north": 0.47}, "solvable": {"false": 0.54, "true": 0.46}, "value": {"0": 0.06, "1": 0.5, "2": 0.24, "3": 0.2}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:5fbead1d9adf0198a1370b5f", "state_id": "grid_navigation:5fbead1d9adf0198a1370b5f", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # . A\n. # . .\nG . # .\n# . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.5, "south": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 3), (1, 2), (1, 3), (2, 0), (2, 1), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [1, 1], [2, 2], [3, 0], [3, 2], [3, 3]], "position": [0, 3], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.56, "west": 0.44}, "solvable": {"false": 0.4, "true": 0.6}, "value": {"0": 0.02, "1": 0.47, "2": 0.32, "3": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:d42273939e836a7050927941", "state_id": "tic_tac_toe:d42273939e836a7050927941", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O O\nO X .\nX X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place O in row 2, column 3 (cell 6).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_6": 0.0, "cell_9": 1.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXXOXOX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOOOX.XX.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.44, "cell_9": 0.56}, "solvable": {"false": 0.85, "true": 0.15}, "value": {"0": 0.53, "1": 0.35, "2": 0.12}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:d417a671adc8817813e42134", "state_id": "grid_navigation:d417a671adc8817813e42134", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG # . #\n# # A .\n. # # .\n. # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 3), (1, 0), (1, 1), (2, 1), (2, 2), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 3], [1, 0], [1, 1], [2, 1], [2, 2], [3, 1]], "position": [1, 2], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.39, "east": 0.61}, "solvable": {"false": 0.74, "true": 0.26}, "value": {"0": 0.15, "1": 0.17, "2": 0.13, "3": 0.55}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:66b4b5872c28fb0fa4c2b953", "state_id": "grid_navigation:66b4b5872c28fb0fa4c2b953", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # #\n# G . #\n. # # .\n# . . A", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 1), (0, 3), (1, 2), (2, 0), (2, 2), (3, 0), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [0, 3], [1, 0], [1, 3], [2, 1], [2, 2], [3, 0]], "position": [3, 3], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.6, "west": 0.4}, "solvable": {"false": 0.45999999999999996, "true": 0.54}, "value": {"0": 0.02, "1": 0.45, "2": 0.3, "3": 0.23}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:7e0a1ba6092e721e3ecf9478", "state_id": "grid_navigation:7e0a1ba6092e721e3ecf9478", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\n. # . #\nA # . .\n. # G .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"south": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 2), (2, 0), (2, 2), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 1], [1, 3], [2, 1], [3, 1]], "position": [2, 0], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.54, "south": 0.46}, "solvable": {"false": 0.49, "true": 0.51}, "value": {"0": 0.03, "1": 0.44, "2": 0.34, "3": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:282936215a7af2ac89275943", "state_id": "grid_navigation:282936215a7af2ac89275943", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . G .\n. # . A\n. . . .\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.5, "west": 0.5, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 1]], "position": [1, 3], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.6, "north": 0.29, "south": 0.11}, "solvable": {"false": 0.22999999999999998, "true": 0.77}, "value": {"0": 0.06, "1": 0.68, "2": 0.2, "3": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:cec4b9b5ed0c25151fcecc67", "state_id": "grid_navigation:cec4b9b5ed0c25151fcecc67", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. . . .\nG # . A\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.3333333333333333, "south": 0.3333333333333333, "west": 0.3333333333333333}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 1),)", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[2, 1]], "position": [2, 3], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.6, "north": 0.15, "south": 0.25}, "solvable": {"false": 0.36, "true": 0.64}, "value": {"0": 0.04, "1": 0.57, "2": 0.35, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:3652194c011889d0987391ba", "state_id": "grid_navigation:3652194c011889d0987391ba", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . G .\n. # # .\n# . # .\n. A # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.5, "west": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 1), (1, 2), (2, 1), (2, 3), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 1], [1, 2], [2, 0], [2, 2], [3, 2]], "position": [3, 1], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.36, "north": 0.64}, "solvable": {"false": 0.52, "true": 0.48}, "value": {"0": 0.02, "1": 0.35, "2": 0.46, "3": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:1b915a9cb3483461b5dd30e4", "state_id": "grid_navigation:1b915a9cb3483461b5dd30e4", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . A .\n. . # .\n# . . G\n# . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 1.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 2), (3, 0), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 2], [2, 0], [3, 0]], "position": [0, 2], "goal": [2, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.15, "east": 0.85}, "solvable": {"false": 0.15000000000000002, "true": 0.85}, "value": {"0": 0.01, "1": 0.64, "2": 0.31, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:4f21018c2a71f0cd392056dc", "state_id": "tic_tac_toe:4f21018c2a71f0cd392056dc", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . X\n. O O\nX X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 1.0, "cell_4": 0.0, "cell_2": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXXO.X.O:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.X.OOXX.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.66, "cell_2": 0.1, "cell_4": 0.24}, "solvable": {"false": 0.28, "true": 0.72}, "value": {"0": 0.08, "1": 0.1, "2": 0.82}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:6073d5b20fbe57491c4380e4", "state_id": "tic_tac_toe:6073d5b20fbe57491c4380e4", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X O\nX . O\nO X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.5, "cell_5": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOX.XOXX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XXOX.OOX.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.47, "cell_9": 0.53}, "solvable": {"false": 0.85, "true": 0.15}, "value": {"0": 0.25, "1": 0.58, "2": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:123f3070f56a1ce96edc494a", "state_id": "tic_tac_toe:123f3070f56a1ce96edc494a", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO O .\n. X X\nX O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place O in row 2, column 1 (cell 4).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_3": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOXX.XOX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OO..XXXOX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.6, "cell_4": 0.4}, "solvable": {"false": 0.84, "true": 0.16}, "value": {"0": 0.52, "1": 0.3, "2": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:84ab36bc52e8794b90b87156", "state_id": "grid_navigation:84ab36bc52e8794b90b87156", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # A .\n# # . .\n. # . .\n# . G .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.0, "south": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 2), (2, 2), (2, 3), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 0], [1, 1], [2, 1], [3, 0]], "position": [0, 2], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.38, "east": 0.62}, "solvable": {"false": 0.31999999999999995, "true": 0.68}, "value": {"0": 0.01, "1": 0.45, "2": 0.42, "3": 0.12}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:93af15ed7f0c36a7c9e4918e", "state_id": "tic_tac_toe:93af15ed7f0c36a7c9e4918e", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O X\nX O O\n. X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_1": "Place X in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_7", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 1.0, "cell_9": 0.0, "cell_1": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXXOO.X.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".OXXOO.X.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.33, "cell_9": 0.31, "cell_7": 0.36}, "solvable": {"false": 0.62, "true": 0.38}, "value": {"0": 0.3, "1": 0.32, "2": 0.38}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:871e8d9985540ed41d339ef0", "state_id": "grid_navigation:871e8d9985540ed41d339ef0", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # #\n# # . .\n. # . #\nA . # G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 3), (1, 0), (1, 3), (2, 1), (2, 2), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [1, 0], [1, 1], [2, 1], [2, 3], [3, 2]], "position": [3, 0], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.75, "north": 0.25}, "solvable": {"false": 0.48, "true": 0.52}, "value": {"0": 0.06, "1": 0.35, "2": 0.29, "3": 0.3}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:44d004ed0571c7e5074d6497", "state_id": "tic_tac_toe:44d004ed0571c7e5074d6497", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X O\nX . O\n. O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place O in row 2, column 2 (cell 5).", "cell_7": "Place O in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_5": 0.5, "cell_7": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXX.OXXO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XXOX.O.OX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.38, "cell_5": 0.62}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.21, "1": 0.58, "2": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:0fb779b2163c4dfb66b2ba48", "state_id": "tic_tac_toe:0fb779b2163c4dfb66b2ba48", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X O\n. X .\nX O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place O in row 2, column 1 (cell 4).", "cell_6": "Place O in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_4": 0.5, "cell_6": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.XXXOO.X:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXO.X.XOX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.27, "cell_4": 0.73}, "solvable": {"false": 0.89, "true": 0.11}, "value": {"0": 0.4, "1": 0.47, "2": 0.13}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:d57488fc61ca625dc30ffcf2", "state_id": "grid_navigation:d57488fc61ca625dc30ffcf2", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . #\n. # . .\n. A . .\nG . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.5, "south": 0.5, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((1, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 1]], "position": [2, 1], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.02, "south": 0.5, "west": 0.48}, "solvable": {"false": 0.16000000000000003, "true": 0.84}, "value": {"0": 0.29, "1": 0.5599999999999999, "2": 0.12, "3": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:477cfbd258c2a348c4dc21ec", "state_id": "tic_tac_toe:477cfbd258c2a348c4dc21ec", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O O\n. X .\nX . X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place O in row 2, column 1 (cell 4).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_6": "Place O in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_1": 1.0, "cell_8": 0.0, "cell_6": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOX.O.X:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".OO.X.X.X", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.3, "cell_8": 0.23, "cell_1": 0.29, "cell_6": 0.18}, "solvable": {"false": 0.83, "true": 0.17}, "value": {"0": 0.45, "1": 0.29, "2": 0.26}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:6a076e2f02af899c042c0df8", "state_id": "grid_navigation:6a076e2f02af899c042c0df8", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . #\n# G A .\n. . . .\n# . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west.", "east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "west": 1.0, "east": 0.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (0, 3), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 0], [3, 0]], "position": [1, 2], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.99, "east": 0, "south": 0, "north": 0.01}, "solvable": {"false": 0.06999999999999995, "true": 0.93}, "value": {"0": 0.95, "1": 0.03, "2": 0.01, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:716bb10d627444f113d60b46", "state_id": "tic_tac_toe:716bb10d627444f113d60b46", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . .\n. O .\nX X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place X in row 2, column 3 (cell 6).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.0, "cell_4": 0.0, "cell_3": 0.0, "cell_2": 0.0, "cell_9": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....OXO.X:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O...O.XX.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.15, "cell_4": 0.11, "cell_2": 0.05, "cell_3": 0.23, "cell_9": 0.46}, "solvable": {"false": 0.38, "true": 0.62}, "value": {"0": 0.11, "1": 0.19, "2": 0.7}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:cedd12b80f62d0dd2a0fd6ce", "state_id": "grid_navigation:cedd12b80f62d0dd2a0fd6ce", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # # .\n. . . #\n# # G .\n. # . A", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.5, "north": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 2), (1, 2), (1, 3), (2, 0), (3, 1), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [0, 2], [1, 3], [2, 0], [2, 1], [3, 1]], "position": [3, 3], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.35, "north": 0.65}, "solvable": {"false": 0.28, "true": 0.72}, "value": {"0": 0.07, "1": 0.57, "2": 0.24, "3": 0.12}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:3c1c2706cf0ce8f43b17a4b0", "state_id": "tic_tac_toe:3c1c2706cf0ce8f43b17a4b0", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X O\nO . .\n. . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place X in row 3, column 2 (cell 8).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_8": 0.5, "cell_1": 0.0, "cell_6": 0.0, "cell_7": 0.0, "cell_5": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XO...XO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XOO....X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.56, "cell_6": 0.12, "cell_8": 0.12, "cell_1": 0.11, "cell_7": 0.09}, "solvable": {"false": 0.78, "true": 0.22}, "value": {"0": 0.21, "1": 0.53, "2": 0.26}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:fb7a217c9e4d87ade2690091", "state_id": "grid_navigation:fb7a217c9e4d87ade2690091", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nA G . #\n. # # #\n. # . .\n. . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 1), (1, 2), (1, 3), (2, 1), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 1], [1, 2], [1, 3], [2, 1], [3, 2]], "position": [0, 0], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.23, "east": 0.77}, "solvable": {"false": 0.6599999999999999, "true": 0.34}, "value": {"0": 0.74, "1": 0.06, "2": 0.05, "3": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:c83a81bbc640b2339d2486a9", "state_id": "grid_navigation:c83a81bbc640b2339d2486a9", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n# # . A\n# # G .\n. # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.5, "north": 0.0, "west": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 2), (1, 2), (1, 3), (2, 2), (2, 3), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 0], [1, 1], [2, 0], [2, 1], [3, 1]], "position": [1, 3], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.15, "west": 0.48, "south": 0.37}, "solvable": {"false": 0.32999999999999996, "true": 0.67}, "value": {"0": 0.18, "1": 0.53, "2": 0.15, "3": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:6ceea6db887683231da45c06", "state_id": "grid_navigation:6ceea6db887683231da45c06", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\n. . # #\nG # # .\n# # . A", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (1, 0), (1, 1), (1, 3), (2, 1), (2, 2), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 2], [1, 3], [2, 1], [2, 2], [3, 0], [3, 1]], "position": [3, 3], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.38, "north": 0.62}, "solvable": {"false": 0.62, "true": 0.38}, "value": {"0": 0.02, "1": 0.25, "2": 0.31, "3": 0.42}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:4f10b34c1cede17f8170ad7b", "state_id": "grid_navigation:4f10b34c1cede17f8170ad7b", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . A G\n# . # .\n# # # .\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((1, 1), (1, 2), (2, 2), (3, 1), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [1, 2], [2, 0], [2, 1], [2, 2]], "position": [0, 2], "goal": [0, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.94, "west": 0.06}, "solvable": {"false": 0.08999999999999997, "true": 0.91}, "value": {"0": 0.9299999999999999, "1": 0.03, "2": 0.02, "3": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:ae621a863ae1c6c53cd59ad8", "state_id": "tic_tac_toe:ae621a863ae1c6c53cd59ad8", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . .\nX . O\n. . O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_8": "Place X in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.5, "cell_5": 0.0, "cell_3": 0.5, "cell_2": 0.0, "cell_8": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OX.OX..:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X..X.O..O", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.38, "cell_3": 0.11, "cell_8": 0.15, "cell_7": 0.17, "cell_2": 0.19}, "solvable": {"false": 0.62, "true": 0.38}, "value": {"0": 0.28, "1": 0.37, "2": 0.35}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:4cad0a02a210411fe343cb1c", "state_id": "grid_navigation:4cad0a02a210411fe343cb1c", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A . .\n# . # #\n. # # #\n# # # G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.3333333333333333, "east": 0.3333333333333333, "south": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 2), (0, 3), (1, 0), (1, 1), (1, 2), (2, 0), (2, 1), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [1, 2], [1, 3], [2, 1], [2, 2], [2, 3], [3, 0], [3, 1], [3, 2]], "position": [0, 1], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.32, "east": 0.47000000000000003, "west": 0.21}, "solvable": {"false": 0.84, "true": 0.16}, "value": {"0": 0.01, "1": 0.1, "2": 0.1, "3": 0.79}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:d5fa1fc484e7caf2be0468a2", "state_id": "grid_navigation:d5fa1fc484e7caf2be0468a2", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . G\n. . # .\n# # # #\n. A . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "west": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (0, 3), (1, 1), (1, 2), (2, 2), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 2], [2, 0], [2, 1], [2, 2], [2, 3], [3, 3]], "position": [3, 1], "goal": [0, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.27, "west": 0.73}, "solvable": {"false": 0.5700000000000001, "true": 0.43}, "value": {"0": 0.03, "1": 0.41000000000000003, "2": 0.25, "3": 0.31}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:610a81b3e3ddadb22b72bb2b", "state_id": "grid_navigation:610a81b3e3ddadb22b72bb2b", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . #\n. G # .\n. A . #\n# # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 1.0, "east": 0.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (0, 3), (1, 2), (2, 3), (3, 0), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 3], [1, 2], [2, 3], [3, 0], [3, 1]], "position": [2, 1], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.06, "west": 0.1, "north": 0.84}, "solvable": {"false": 0.30000000000000004, "true": 0.7}, "value": {"0": 0.63, "1": 0.19, "2": 0.07, "3": 0.11}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:ab2392ca2db90c4ebd59b736", "state_id": "grid_navigation:ab2392ca2db90c4ebd59b736", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . G .\n. . # .\n# # # #\n. A . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "west": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 1), (2, 0), (2, 1), (2, 2), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 2], [2, 0], [2, 1], [2, 2], [2, 3]], "position": [3, 1], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.64, "west": 0.36}, "solvable": {"false": 0.51, "true": 0.49}, "value": {"0": 0.03, "1": 0.35, "2": 0.29, "3": 0.33}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:71c0dfca7c64c1ca590525e2", "state_id": "grid_navigation:71c0dfca7c64c1ca590525e2", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . .\n# # # A\n. # # .\n# G # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.5, "south": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 2), (0, 3), (1, 1), (1, 2), (2, 0), (2, 1), (2, 2), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 0], [1, 1], [1, 2], [2, 1], [2, 2], [3, 0], [3, 2], [3, 3]], "position": [1, 3], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.38, "south": 0.62}, "solvable": {"false": 0.69, "true": 0.31}, "value": {"0": 0.03, "1": 0.29, "2": 0.2, "3": 0.48}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:9136e3750bbf17939a69f54a", "state_id": "tic_tac_toe:9136e3750bbf17939a69f54a", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . O\nO X .\nX . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_8", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.0, "cell_6": 0.0, "cell_8": 0.5, "cell_2": 0.0, "cell_9": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O.X.XO.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..OOX.X..", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.23, "cell_2": 0.19, "cell_6": 0.2, "cell_1": 0.17, "cell_9": 0.21}, "solvable": {"false": 0.75, "true": 0.25}, "value": {"0": 0.28, "1": 0.41, "2": 0.31}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:63eeeb4f94a1c368fae77d5d", "state_id": "grid_navigation:63eeeb4f94a1c368fae77d5d", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # #\n. # # .\nA G # #\n. # . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 1.0, "south": 0.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (0, 3), (1, 2), (1, 3), (2, 1), (2, 2), (3, 1), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [0, 3], [1, 1], [1, 2], [2, 2], [2, 3], [3, 1], [3, 3]], "position": [2, 0], "goal": [2, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.16, "south": 0.1, "east": 0.74}, "solvable": {"false": 0.09999999999999998, "true": 0.9}, "value": {"0": 0.98, "1": 0.01, "2": 0, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:7ff709da932a16d27d1d997a", "state_id": "tic_tac_toe:7ff709da932a16d27d1d997a", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X .\n. O .\n. . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place X in row 3, column 2 (cell 8).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_4": "Place X in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_8": 0.0, "cell_6": 0.25, "cell_3": 0.25, "cell_7": 0.25, "cell_4": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_4", "cell_6", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O.OXX..:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OX..O...X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.27999999999999997, "cell_6": 0.14, "cell_7": 0.21, "cell_8": 0.18, "cell_4": 0.19}, "solvable": {"false": 0.6799999999999999, "true": 0.32}, "value": {"0": 0.13, "1": 0.43, "2": 0.44}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:750aafb2392767f83388d37a", "state_id": "grid_navigation:750aafb2392767f83388d37a", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n# . # .\n. G . A\n. . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 2), (2, 2), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [1, 2], [3, 3]], "position": [2, 3], "goal": [2, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.12, "west": 0.88}, "solvable": {"false": 0.16000000000000003, "true": 0.84}, "value": {"0": 0.42, "1": 0.43, "2": 0.11, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:4637206188779421325d1b11", "state_id": "grid_navigation:4637206188779421325d1b11", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A . .\n. . # #\nG . # .\n# # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 0.5, "south": 0.5, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (1, 0), (2, 0), (2, 1), (2, 2), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 2], [1, 3], [2, 2], [3, 0], [3, 1], [3, 2]], "position": [0, 1], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.01, "south": 0.23, "west": 0.76}, "solvable": {"false": 0.16000000000000003, "true": 0.84}, "value": {"0": 0.19, "1": 0.64, "2": 0.11, "3": 0.06}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:552710f0ef25834c4c41cbd3", "state_id": "grid_navigation:552710f0ef25834c4c41cbd3", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # . #\n. G A .\n# . # #\n# # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 1.0, "north": 0.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 1), (0, 3), (2, 0), (2, 2), (2, 3), (3, 0), (3, 1), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [0, 3], [2, 0], [2, 2], [2, 3], [3, 0], [3, 1], [3, 2]], "position": [1, 2], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.02, "east": 0, "west": 0.98}, "solvable": {"false": 0.08999999999999997, "true": 0.91}, "value": {"0": 0.95, "1": 0.03, "2": 0.01, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:6471ddd16d38abf528d602cb", "state_id": "grid_navigation:6471ddd16d38abf528d602cb", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. # . .\n. . G A\n# . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "west": 1.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 3), (2, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 1], [3, 0]], "position": [2, 3], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.02, "south": 0.01, "west": 0.97}, "solvable": {"false": 0.050000000000000044, "true": 0.95}, "value": {"0": 0.96, "1": 0.03, "2": 0.01, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:384c9456106a304cad929aa0", "state_id": "grid_navigation:384c9456106a304cad929aa0", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# G # .\n# . . .\n. # # A\n# # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 2), (1, 0), (2, 1), (2, 2), (3, 0), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [1, 0], [2, 1], [2, 2], [3, 0], [3, 1]], "position": [2, 3], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.88, "south": 0.12}, "solvable": {"false": 0.39, "true": 0.61}, "value": {"0": 0.02, "1": 0.38, "2": 0.44, "3": 0.16}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:208420c4b85344f36483d956", "state_id": "tic_tac_toe:208420c4b85344f36483d956", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O O\nX X .\n. . O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place X in row 2, column 3 (cell 6).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_8": "Place X in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_6", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.5, "cell_7": 0.5, "cell_8": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_6", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXX.XOO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOOXX...O", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.68, "cell_8": 0.16, "cell_7": 0.16}, "solvable": {"false": 0.53, "true": 0.47}, "value": {"0": 0.14, "1": 0.37, "2": 0.49}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:903de5f983cb5adf178bff3c", "state_id": "tic_tac_toe:903de5f983cb5adf178bff3c", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . X\nX . O\nO O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place O in row 1, column 2 (cell 2).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_2": 0.5, "cell_5": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:OOXX.OX.X:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.XX.OOOX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.63, "cell_2": 0.37}, "solvable": {"false": 0.78, "true": 0.22}, "value": {"0": 0.32, "1": 0.34, "2": 0.34}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:77b18ceb019d1bae4e18e715", "state_id": "grid_navigation:77b18ceb019d1bae4e18e715", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . G #\n. . # .\n# A . .\n. # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (1, 1), (2, 3), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 2], [2, 0], [3, 1]], "position": [2, 1], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.24, "north": 0.76}, "solvable": {"false": 0.39, "true": 0.61}, "value": {"0": 0.04, "1": 0.54, "2": 0.26, "3": 0.16}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:bbd0e805eacf2de01095769c", "state_id": "tic_tac_toe:bbd0e805eacf2de01095769c", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O O\n. X X\nO X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.5, "cell_4": 0.5, "cell_9": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_4"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OOXXOX.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".OO.XXOX.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.25, "cell_9": 0.53, "cell_4": 0.22}, "solvable": {"false": 0.45999999999999996, "true": 0.54}, "value": {"0": 0.2, "1": 0.3, "2": 0.5}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:47294a3d7ce1c36a45bdc180", "state_id": "tic_tac_toe:47294a3d7ce1c36a45bdc180", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X .\n. O O\nX X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_3": "Place X in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 1.0, "cell_4": 0.0, "cell_3": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.O.XOXO.X:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OX..OOXX.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.27, "cell_9": 0.46, "cell_4": 0.27}, "solvable": {"false": 0.29000000000000004, "true": 0.71}, "value": {"0": 0.09, "1": 0.14, "2": 0.77}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:62904f9e38892480fd63f01f", "state_id": "tic_tac_toe:62904f9e38892480fd63f01f", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X X\nO . O\n. . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place X in row 3, column 2 (cell 8).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_8": 0.3333333333333333, "cell_7": 0.3333333333333333, "cell_5": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5", "cell_7", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XO.OOXX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXXO.O..X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.55, "cell_8": 0.23, "cell_7": 0.22}, "solvable": {"false": 0.6599999999999999, "true": 0.34}, "value": {"0": 0.26, "1": 0.37, "2": 0.37}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:e79441974f9f61292d03b3ca", "state_id": "grid_navigation:e79441974f9f61292d03b3ca", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # # .\n. # # #\n. . # #\n. A G .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "west": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (1, 1), (2, 0), (2, 1), (2, 2), (3, 1), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [0, 2], [1, 1], [1, 2], [1, 3], [2, 2], [2, 3]], "position": [3, 1], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.95, "west": 0.01, "north": 0.04}, "solvable": {"false": 0.08999999999999997, "true": 0.91}, "value": {"0": 0.97, "1": 0.01, "2": 0.01, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:6fdeaffbf0f7a2da5758dae2", "state_id": "tic_tac_toe:6fdeaffbf0f7a2da5758dae2", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . .\nX O O\nX X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place X in row 1, column 2 (cell 2).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_3": "Place X in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.0, "cell_9": 1.0, "cell_3": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OOOX.XX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O..XOOXX.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.26, "cell_2": 0.12, "cell_9": 0.62}, "solvable": {"false": 0.37, "true": 0.63}, "value": {"0": 0.11, "1": 0.14, "2": 0.75}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:e568d9de3675211b1305f4bf", "state_id": "grid_navigation:e568d9de3675211b1305f4bf", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . .\n# # A G\n. . . .\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "east": 1.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((2, 1), (3, 0), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 0], [1, 1]], "position": [1, 2], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0, "east": 1, "south": 0}, "solvable": {"false": 0.040000000000000036, "true": 0.96}, "value": {"0": 0.99, "1": 0, "2": 0, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:fdcc9c79531ac5427b797f6d", "state_id": "tic_tac_toe:fdcc9c79531ac5427b797f6d", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X O\nX X .\nO . X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place O in row 2, column 3 (cell 6).", "cell_8": "Place O in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_6", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_6": 0.5, "cell_8": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_6", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.XXX.OXO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXOXX.O.X", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.54, "cell_8": 0.46}, "solvable": {"false": 0.86, "true": 0.14}, "value": {"0": 0.4, "1": 0.44, "2": 0.16}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:e74a03b9ebcbc17f6a632b16", "state_id": "tic_tac_toe:e74a03b9ebcbc17f6a632b16", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O .\n. X X\nO X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place O in row 2, column 1 (cell 4).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_4": 1.0, "cell_3": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXXX.OXO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XO..XXOXO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.38, "cell_3": 0.62}, "solvable": {"false": 0.9, "true": 0.1}, "value": {"0": 0.6, "1": 0.26, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:00d988d8b556a13af5ac0f9d", "state_id": "tic_tac_toe:00d988d8b556a13af5ac0f9d", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . .\n. O X\nX O O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.5, "cell_3": 0.0, "cell_2": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_4"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XXO.OOX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X...OXXOO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.31, "cell_4": 0.4, "cell_2": 0.29}, "solvable": {"false": 0.6799999999999999, "true": 0.32}, "value": {"0": 0.28, "1": 0.34, "2": 0.38}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:bc790e55c39b631bfff94d94", "state_id": "grid_navigation:bc790e55c39b631bfff94d94", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . A\n. # # .\n. G . #\n. # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.5, "south": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (1, 3), (2, 1), (2, 2), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 1], [1, 2], [2, 3], [3, 1]], "position": [0, 3], "goal": [2, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.63, "south": 0.37}, "solvable": {"false": 0.33999999999999997, "true": 0.66}, "value": {"0": 0.03, "1": 0.53, "2": 0.31, "3": 0.13}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:d63b86a7604f5cb0872f1e5a", "state_id": "grid_navigation:d63b86a7604f5cb0872f1e5a", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . A\n# . # .\nG . . .\n# . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 0.5, "south": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 2), (0, 3), (2, 0), (2, 2), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 0], [1, 2], [3, 0], [3, 2], [3, 3]], "position": [0, 3], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.47, "south": 0.53}, "solvable": {"false": 0.21999999999999997, "true": 0.78}, "value": {"0": 0.02, "1": 0.65, "2": 0.24, "3": 0.09}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:f3195b3cc3dd5881b89cf484", "state_id": "tic_tac_toe:f3195b3cc3dd5881b89cf484", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X .\nX O .\nX O O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place X in row 2, column 3 (cell 6).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_3": "Place X in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.0, "cell_1": 1.0, "cell_3": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXOO.XX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".X.XO.XOO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.43, "cell_6": 0.24, "cell_3": 0.33}, "solvable": {"false": 0.52, "true": 0.48}, "value": {"0": 0.13, "1": 0.19, "2": 0.68}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:742abe2d9fe3944d39b12dc7", "state_id": "tic_tac_toe:742abe2d9fe3944d39b12dc7", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . O\nX O .\nX O X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 1.0, "cell_2": 0.0, "cell_6": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXO.XOX:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..OXO.XOX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.35, "cell_2": 0.18, "cell_6": 0.47}, "solvable": {"false": 0.6599999999999999, "true": 0.34}, "value": {"0": 0.34, "1": 0.35, "2": 0.31}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:566ce0830634374aad702ded", "state_id": "grid_navigation:566ce0830634374aad702ded", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # #\n. A . .\nG . # .\n# . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east.", "north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.5, "east": 0.0, "north": 0.0, "south": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 3), (2, 1), (2, 3), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [0, 3], [2, 2], [3, 0]], "position": [1, 1], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.37, "west": 0.61, "north": 0.01, "east": 0.01}, "solvable": {"false": 0.14, "true": 0.86}, "value": {"0": 0.56, "1": 0.36, "2": 0.05, "3": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:01f22728d4138f9c0af7258d", "state_id": "grid_navigation:01f22728d4138f9c0af7258d", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG . # #\n. . A .\n. # . .\n# . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 1.0, "east": 0.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (0, 3), (2, 1), (3, 0), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [2, 1], [3, 0], [3, 2]], "position": [1, 2], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.24, "west": 0.67, "south": 0.09}, "solvable": {"false": 0.30000000000000004, "true": 0.7}, "value": {"0": 0.18, "1": 0.48, "2": 0.16, "3": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:4dd1df8bc2e55cd12eecf193", "state_id": "grid_navigation:4dd1df8bc2e55cd12eecf193", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. . . .\nG . A #\n. . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 1.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((2, 0), (3, 0), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[2, 3], [3, 2], [3, 3]], "position": [2, 2], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.03, "west": 0.97}, "solvable": {"false": 0.12, "true": 0.88}, "value": {"0": 0.29, "1": 0.48, "2": 0.18, "3": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:bcac1640cac78f9f6ba52dad", "state_id": "tic_tac_toe:bcac1640cac78f9f6ba52dad", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O X\nO X .\n. X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_7", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.0, "cell_6": 0.0, "cell_7": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.O.OXXX.O:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".OXOX..XO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.42, "cell_1": 0.28, "cell_7": 0.3}, "solvable": {"false": 0.6599999999999999, "true": 0.34}, "value": {"0": 0.09, "1": 0.45, "2": 0.46}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:5b25c51c87cd9beea6522571", "state_id": "grid_navigation:5b25c51c87cd9beea6522571", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # .\nG . # .\n# # . A\n# . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"south": 0.3333333333333333, "west": 0.3333333333333333, "north": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 2), (0, 3), (1, 2), (2, 0), (2, 1), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [1, 2], [2, 0], [2, 1], [3, 0], [3, 2]], "position": [2, 3], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.59, "south": 0.08, "north": 0.33}, "solvable": {"false": 0.39, "true": 0.61}, "value": {"0": 0.02, "1": 0.45, "2": 0.33, "3": 0.2}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:0bdfb1843c9fe2f8060025e4", "state_id": "grid_navigation:0bdfb1843c9fe2f8060025e4", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. . A .\n# . # #\n. # G #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.3333333333333333, "west": 0.3333333333333333, "east": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 2), (1, 0), (1, 1), (1, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[2, 0], [2, 2], [2, 3], [3, 1], [3, 3]], "position": [1, 2], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.73, "north": 0.17, "east": 0.1}, "solvable": {"false": 0.47, "true": 0.53}, "value": {"0": 0.08, "1": 0.48, "2": 0.19, "3": 0.25}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:f161266fc3d91acc297555ec", "state_id": "tic_tac_toe:f161266fc3d91acc297555ec", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X O\n. X .\nX O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_4": "Place X in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_9": 0.3333333333333333, "cell_6": 0.3333333333333333, "cell_4": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4", "cell_6", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OOXXX.O:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXO.X.XO.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.19, "cell_4": 0.35, "cell_9": 0.46}, "solvable": {"false": 0.56, "true": 0.44}, "value": {"0": 0.12, "1": 0.3, "2": 0.58}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:1f8620ffe8962675da19ab6e", "state_id": "grid_navigation:1f8620ffe8962675da19ab6e", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # G #\n# # . #\n. . # .\n# . A .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "west": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 2), (1, 0), (1, 2), (1, 3), (2, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 3], [1, 0], [1, 1], [1, 3], [2, 2], [3, 0]], "position": [3, 2], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.35, "west": 0.65}, "solvable": {"false": 0.5800000000000001, "true": 0.42}, "value": {"0": 0.01, "1": 0.23, "2": 0.34, "3": 0.42}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:8081f0314cc646eb77aa73e2", "state_id": "grid_navigation:8081f0314cc646eb77aa73e2", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # .\n. # . #\n# . G #\n. A # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (0, 2), (1, 0), (1, 3), (2, 2), (2, 3), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [1, 1], [1, 3], [2, 0], [2, 3], [3, 2]], "position": [3, 1], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.75, "west": 0.25}, "solvable": {"false": 0.41000000000000003, "true": 0.59}, "value": {"0": 0.34, "1": 0.41, "2": 0.11, "3": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:585958649127100b7faa1c4f", "state_id": "tic_tac_toe:585958649127100b7faa1c4f", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O O\nO X X\n. . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place X in row 3, column 2 (cell 8).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_8": 0.0, "cell_9": 1.0, "cell_7": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...OXXXOO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOOOXX...", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.59, "cell_9": 0.22, "cell_8": 0.19}, "solvable": {"false": 0.73, "true": 0.27}, "value": {"0": 0.2, "1": 0.5, "2": 0.3}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:09e44d43e5d641d78764a77e", "state_id": "grid_navigation:09e44d43e5d641d78764a77e", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # # .\n. # . .\n# # # A\nG # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.5, "south": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 2), (1, 0), (1, 1), (1, 2), (2, 1), (3, 0), (3, 1), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [0, 2], [1, 1], [2, 0], [2, 1], [2, 2], [3, 1], [3, 2]], "position": [2, 3], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.51, "south": 0.49}, "solvable": {"false": 0.6, "true": 0.4}, "value": {"0": 0.04, "1": 0.28, "2": 0.32, "3": 0.36}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:4557f4b9b976ff0f101db817", "state_id": "grid_navigation:4557f4b9b976ff0f101db817", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG . . A\n. # # .\n. # # #\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((1, 1), (1, 2), (2, 1), (2, 2), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 1], [1, 2], [2, 1], [2, 2], [2, 3]], "position": [0, 3], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.38, "west": 0.62}, "solvable": {"false": 0.41000000000000003, "true": 0.59}, "value": {"0": 0.19, "1": 0.42, "2": 0.3, "3": 0.09}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:a41e17cf808c58553b9349ed", "state_id": "grid_navigation:a41e17cf808c58553b9349ed", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . #\n# # . #\n. . . #\nA G . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 3), (2, 0), (2, 1), (2, 3), (3, 1), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 3], [1, 0], [1, 1], [1, 3], [2, 3]], "position": [3, 0], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.58, "north": 0.42}, "solvable": {"false": 0.14, "true": 0.86}, "value": {"0": 0.99, "1": 0.01, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:4f4a33c3a6899a2cbf7fcd99", "state_id": "tic_tac_toe:4f4a33c3a6899a2cbf7fcd99", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . X\nO O X\n. . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place X in row 1, column 2 (cell 2).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_1": "Place X in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.3333333333333333, "cell_8": 0.0, "cell_9": 0.3333333333333333, "cell_7": 0.0, "cell_1": 0.3333333333333333}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_2", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...OOX..X:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..XOOX...", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.14, "cell_2": 0.18, "cell_9": 0.26, "cell_1": 0.22, "cell_7": 0.2}, "solvable": {"false": 0.5700000000000001, "true": 0.43}, "value": {"0": 0.15, "1": 0.32, "2": 0.53}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:937282d856f55408a08fa6b5", "state_id": "tic_tac_toe:937282d856f55408a08fa6b5", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . X\nX O .\n. . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place X in row 2, column 3 (cell 6).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_6": 0.0, "cell_9": 1.0, "cell_7": 0.0, "cell_8": 0.0, "cell_2": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....OXX.O:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.XXO....", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.15, "cell_6": 0.27, "cell_9": 0.24, "cell_8": 0.12, "cell_7": 0.22}, "solvable": {"false": 0.8, "true": 0.2}, "value": {"0": 0.18, "1": 0.48, "2": 0.34}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:7b793dcd0ade9dd0b6ca172c", "state_id": "tic_tac_toe:7b793dcd0ade9dd0b6ca172c", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X .\n. O .\nX O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place O in row 2, column 1 (cell 4).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_1": "Place O in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_4": 0.25, "cell_3": 0.25, "cell_6": 0.25, "cell_1": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_3", "cell_4", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XXOO..X:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".X..O.XOX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.31, "cell_3": 0.15, "cell_1": 0.33, "cell_6": 0.21}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.41, "1": 0.39, "2": 0.2}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:4ff523a1b0dcf7cb22897bb7", "state_id": "grid_navigation:4ff523a1b0dcf7cb22897bb7", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # #\n. A . .\n. # . .\nG . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (1, 1), (3, 1), (3, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [0, 3], [2, 1], [3, 2]], "position": [1, 1], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.8, "east": 0.2}, "solvable": {"false": 0.31000000000000005, "true": 0.69}, "value": {"0": 0.04, "1": 0.55, "2": 0.31, "3": 0.1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:08938a7224db51f8570f1902", "state_id": "tic_tac_toe:08938a7224db51f8570f1902", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . X\nO O .\nX . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place X in row 2, column 3 (cell 6).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_8": "Place X in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_6", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.5, "cell_2": 0.0, "cell_8": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_6", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.XOO.X.X:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.XOO.X.X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.49, "cell_2": 0.2, "cell_8": 0.31}, "solvable": {"false": 0.52, "true": 0.48}, "value": {"0": 0.18, "1": 0.32, "2": 0.5}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:da6f98322d0f5fe0e593da0b", "state_id": "tic_tac_toe:da6f98322d0f5fe0e593da0b", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . .\nO X .\nX O .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place O in row 2, column 3 (cell 6).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_6": 0.25, "cell_2": 0.25, "cell_3": 0.25, "cell_9": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_3", "cell_6", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....XOXOX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X..OX.XO.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.25, "cell_2": 0.16, "cell_3": 0.21, "cell_9": 0.38}, "solvable": {"false": 0.85, "true": 0.15}, "value": {"0": 0.43, "1": 0.34, "2": 0.23}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:8a9d2b251b0e1fa793b28079", "state_id": "tic_tac_toe:8a9d2b251b0e1fa793b28079", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X .\nO X O\nX O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_1": 0.5, "cell_3": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXXXO.OX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".X.OXOXOX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.3, "cell_1": 0.7}, "solvable": {"false": 0.9, "true": 0.1}, "value": {"0": 0.46, "1": 0.44, "2": 0.1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:ee6060bd1b5638d8b04933e0", "state_id": "tic_tac_toe:ee6060bd1b5638d8b04933e0", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\nO X X\nO X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_1": "Place O in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.0, "cell_3": 0.0, "cell_2": 0.0, "cell_1": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...OXXOX.:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "...OXXOX.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.13, "cell_9": 0.53, "cell_2": 0.12, "cell_3": 0.22}, "solvable": {"false": 0.85, "true": 0.15}, "value": {"0": 0.52, "1": 0.29, "2": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:fd632828d1bf6b7aeac27be0", "state_id": "tic_tac_toe:fd632828d1bf6b7aeac27be0", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O X\nO O .\nX X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place X in row 2, column 3 (cell 6).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.0, "cell_1": 0.0, "cell_9": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XXOOXO.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".OXOO.XX.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.35, "cell_9": 0.42, "cell_1": 0.23}, "solvable": {"false": 0.43000000000000005, "true": 0.57}, "value": {"0": 0.23, "1": 0.22, "2": 0.55}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:594041215a2202655649c2ba", "state_id": "grid_navigation:594041215a2202655649c2ba", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. A . #\n. . . G\n# . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west.", "south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.0, "west": 0.0, "south": 0.5, "east": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 0), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 3], [3, 0], [3, 2]], "position": [1, 1], "goal": [2, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.07, "south": 0.79, "north": 0.04, "east": 0.1}, "solvable": {"false": 0.10999999999999999, "true": 0.89}, "value": {"0": 0.52, "1": 0.42, "2": 0.03, "3": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:fbd39b4f261cb2df003fc56b", "state_id": "grid_navigation:fbd39b4f261cb2df003fc56b", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . .\n. . . #\n# . # .\nG A . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 1.0, "north": 0.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 3), (2, 1), (3, 0), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 3], [2, 0], [2, 2], [3, 3]], "position": [3, 1], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.9299999999999999, "north": 0.05, "east": 0.02}, "solvable": {"false": 0.17000000000000004, "true": 0.83}, "value": {"0": 0.74, "1": 0.14, "2": 0.06, "3": 0.06}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:6d0b1a544c25e3062b93f66f", "state_id": "tic_tac_toe:6d0b1a544c25e3062b93f66f", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X O\nO . X\nX O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_1": 0.5, "cell_5": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXX.OOXX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XOO.XXOX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.33, "cell_5": 0.67}, "solvable": {"false": 0.88, "true": 0.12}, "value": {"0": 0.43, "1": 0.38, "2": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:983aa66daf28d22614545cbc", "state_id": "tic_tac_toe:983aa66daf28d22614545cbc", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X X\n. . O\n. X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place O in row 2, column 2 (cell 5).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 1.0, "cell_4": 0.0, "cell_7": 0.0, "cell_9": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OX.X.OX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXX..O.X.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.15, "cell_9": 0.2, "cell_4": 0.18, "cell_5": 0.47}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.47, "1": 0.33, "2": 0.2}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:8aeaf2d990bc25648a345580", "state_id": "grid_navigation:8aeaf2d990bc25648a345580", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n. # # .\n. . . A\nG . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 1.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 3), (2, 1), (2, 2), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 1], [1, 2], [3, 3]], "position": [2, 3], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.18, "west": 0.82}, "solvable": {"false": 0.20999999999999996, "true": 0.79}, "value": {"0": 0.03, "1": 0.64, "2": 0.24, "3": 0.09}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:2c0ca4330f438535f0d868a9", "state_id": "grid_navigation:2c0ca4330f438535f0d868a9", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . G\n. . . #\n. . A .\n# . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east.", "west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 1.0, "east": 0.0, "west": 0.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 3), (3, 0), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 3], [3, 0], [3, 3]], "position": [2, 2], "goal": [0, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.81, "south": 0.01, "east": 0.06, "west": 0.12}, "solvable": {"false": 0.18999999999999995, "true": 0.81}, "value": {"0": 0.04, "1": 0.67, "2": 0.2, "3": 0.09}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:ec22afe81d8bdc9d5def6436", "state_id": "tic_tac_toe:ec22afe81d8bdc9d5def6436", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O .\n. X O\n. X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_7": 0.0, "cell_9": 1.0, "cell_4": 0.0, "cell_3": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XXXO.O.:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XO..XO.X.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.33, "cell_3": 0.2, "cell_4": 0.2, "cell_7": 0.27}, "solvable": {"false": 0.88, "true": 0.12}, "value": {"0": 0.61, "1": 0.24, "2": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:f03371f2dcfe2acb929f28d3", "state_id": "grid_navigation:f03371f2dcfe2acb929f28d3", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. G . .\n# . # #\n. # A .\n# # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "south": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (1, 2), (1, 3), (2, 1), (3, 0), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [1, 2], [1, 3], [2, 1], [3, 0], [3, 1]], "position": [2, 2], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.64, "south": 0.36}, "solvable": {"false": 0.55, "true": 0.45}, "value": {"0": 0.06, "1": 0.49, "2": 0.27, "3": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:89b203dffe925e4e639409ab", "state_id": "tic_tac_toe:89b203dffe925e4e639409ab", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X O\nX O O\n. X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_1": 0.3333333333333333, "cell_7": 0.3333333333333333, "cell_9": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_7", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOXOX.X.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XOXOO.X.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.34, "cell_7": 0.35, "cell_9": 0.31}, "solvable": {"false": 0.6699999999999999, "true": 0.33}, "value": {"0": 0.34, "1": 0.35, "2": 0.31}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:5c3e07c56bc6e24af576105f", "state_id": "grid_navigation:5c3e07c56bc6e24af576105f", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # .\n# . . .\n. # . .\nA . # G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.5, "east": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (1, 2), (2, 3), (3, 1), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [1, 0], [2, 1], [3, 2]], "position": [3, 0], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.62, "north": 0.38}, "solvable": {"false": 0.29000000000000004, "true": 0.71}, "value": {"0": 0.04, "1": 0.43, "2": 0.35, "3": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:c9168ff330595f271fe96fa8", "state_id": "grid_navigation:c9168ff330595f271fe96fa8", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. . # #\nA . . .\n. G . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.5, "north": 0.0, "south": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((2, 2), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 2], [1, 3]], "position": [2, 0], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.06, "south": 0.64, "east": 0.3}, "solvable": {"false": 0.12, "true": 0.88}, "value": {"0": 0.46, "1": 0.46, "2": 0.07, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:faacd94c777ff2b5fb9fb147", "state_id": "tic_tac_toe:faacd94c777ff2b5fb9fb147", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X .\n. X .\n. O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place O in row 1, column 3 (cell 3).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_4": "Place O in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_3": 0.25, "cell_6": 0.25, "cell_7": 0.25, "cell_4": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_4", "cell_6", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OOXXX..:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OX..X..OX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.24, "cell_7": 0.24, "cell_4": 0.35, "cell_6": 0.17}, "solvable": {"false": 0.92, "true": 0.08}, "value": {"0": 0.66, "1": 0.24, "2": 0.1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:145ec62d08f466c690e64d2c", "state_id": "grid_navigation:145ec62d08f466c690e64d2c", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . G\n. # . A\n# . # #\n. . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (1, 2), (1, 3), (2, 1), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 1], [2, 0], [2, 2], [2, 3], [3, 2]], "position": [1, 3], "goal": [0, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.72, "west": 0.28}, "solvable": {"false": 0.33999999999999997, "true": 0.66}, "value": {"0": 0.47, "1": 0.34, "2": 0.12, "3": 0.07}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:c044cad1f9d0b22fd05916f2", "state_id": "tic_tac_toe:c044cad1f9d0b22fd05916f2", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . X\nO . X\n. O .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place O in row 2, column 2 (cell 5).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_5": 0.25, "cell_7": 0.25, "cell_2": 0.25, "cell_9": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_5", "cell_7", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.O.O.XX.X:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.XO.X.O.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.12, "cell_5": 0.45, "cell_7": 0.18, "cell_2": 0.25}, "solvable": {"false": 0.86, "true": 0.14}, "value": {"0": 0.47, "1": 0.34, "2": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:ce8742c6c1b3035fdb09d82a", "state_id": "grid_navigation:ce8742c6c1b3035fdb09d82a", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. . . .\n. # . G\nA . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((2, 1),)", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[2, 1]], "position": [3, 0], "goal": [2, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.39, "east": 0.61}, "solvable": {"false": 0.07999999999999996, "true": 0.92}, "value": {"0": 0.01, "1": 0.64, "2": 0.32, "3": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:640d00da2515c96b6fff2af3", "state_id": "grid_navigation:640d00da2515c96b6fff2af3", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . A #\n# . . .\n. . # #\n# . G .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.5, "west": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 0), (1, 1), (2, 3), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 0], [2, 2], [2, 3], [3, 0]], "position": [0, 2], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.29, "south": 0.71}, "solvable": {"false": 0.26, "true": 0.74}, "value": {"0": 0.01, "1": 0.59, "2": 0.29, "3": 0.11}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:25847604128a17b2465b7af1", "state_id": "grid_navigation:25847604128a17b2465b7af1", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . .\n. A . .\n# . . .\n. . # G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north.", "west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.5, "north": 0.0, "west": 0.0, "east": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (1, 3), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [2, 0], [3, 2]], "position": [1, 1], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.66, "south": 0.29, "west": 0.03, "north": 0.02}, "solvable": {"false": 0.18999999999999995, "true": 0.81}, "value": {"0": 0.01, "1": 0.53, "2": 0.38, "3": 0.08}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:a7f6a79362db14b0e70673be", "state_id": "grid_navigation:a7f6a79362db14b0e70673be", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n. # A G\n# . . .\n. . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 0), (2, 1), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 1], [2, 0], [3, 3]], "position": [1, 2], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0, "east": 1}, "solvable": {"false": 0.050000000000000044, "true": 0.95}, "value": {"0": 0.99, "1": 0, "2": 0, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:33613fd3933be526c8f1b15b", "state_id": "tic_tac_toe:33613fd3933be526c8f1b15b", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . X\nX . O\n. X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_1": 0.25, "cell_2": 0.25, "cell_7": 0.25, "cell_5": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_2", "cell_5", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XX.O.XO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..XX.O.XO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.16, "cell_2": 0.23, "cell_7": 0.17, "cell_5": 0.44}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.35, "1": 0.45, "2": 0.2}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:3ea91dacd018da66c4cc948c", "state_id": "grid_navigation:3ea91dacd018da66c4cc948c", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # # .\n# . # #\n. A . .\nG . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north.", "east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.5, "north": 0.0, "east": 0.0, "west": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (0, 3), (1, 3), (2, 0), (2, 2), (2, 3), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [0, 2], [1, 0], [1, 2], [1, 3], [3, 2]], "position": [2, 1], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.54, "east": 0.01, "north": 0.01, "south": 0.44}, "solvable": {"false": 0.18999999999999995, "true": 0.81}, "value": {"0": 0.38, "1": 0.46, "2": 0.12, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:36a213d8422a97fac626830d", "state_id": "tic_tac_toe:36a213d8422a97fac626830d", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . O\n. . O\nX X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.0, "cell_9": 1.0, "cell_4": 0.0, "cell_5": 0.0, "cell_2": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O..OXX.:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..O..OXX.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.08, "cell_5": 0.2, "cell_2": 0.07, "cell_9": 0.6, "cell_1": 0.05}, "solvable": {"false": 0.43000000000000005, "true": 0.57}, "value": {"0": 0.18, "1": 0.2, "2": 0.62}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:32a179dfe193ef77a2656425", "state_id": "grid_navigation:32a179dfe193ef77a2656425", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . .\n. # # .\n. # G A\n# . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "south": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 3), (1, 2), (2, 1), (2, 2), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 1], [1, 2], [2, 1], [3, 0]], "position": [2, 3], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.12, "west": 0.85, "south": 0.03}, "solvable": {"false": 0.24, "true": 0.76}, "value": {"0": 0.79, "1": 0.11, "2": 0.04, "3": 0.06}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:339ca66874379289daf90548", "state_id": "grid_navigation:339ca66874379289daf90548", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . #\n# # G .\n# # A #\n. # . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 2), (0, 3), (2, 1), (2, 2), (2, 3), (3, 1), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 0], [1, 1], [2, 0], [2, 1], [2, 3], [3, 1], [3, 3]], "position": [2, 2], "goal": [1, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.89, "south": 0.11}, "solvable": {"false": 0.48, "true": 0.52}, "value": {"0": 0.53, "1": 0.11, "2": 0.06, "3": 0.3}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:7cbd5718b9923a2a35f80574", "state_id": "grid_navigation:7cbd5718b9923a2a35f80574", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # .\n. . # .\n. . G .\n. A # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 1.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (2, 1), (3, 1), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [1, 2], [3, 2]], "position": [3, 1], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.21, "north": 0.79}, "solvable": {"false": 0.17000000000000004, "true": 0.83}, "value": {"0": 0.38, "1": 0.49, "2": 0.09, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:5669f7ac5bb3d6b6f0af850d", "state_id": "grid_navigation:5669f7ac5bb3d6b6f0af850d", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG . # .\n. . # #\n. A # .\n. # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.5, "west": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (1, 2), (1, 3), (2, 2), (3, 1), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 2], [1, 3], [2, 2], [3, 1], [3, 2]], "position": [2, 1], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.44, "north": 0.56}, "solvable": {"false": 0.42000000000000004, "true": 0.58}, "value": {"0": 0.1, "1": 0.48, "2": 0.16, "3": 0.26}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:68c42b9c356bd284f6e84422", "state_id": "tic_tac_toe:68c42b9c356bd284f6e84422", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO O X\nX . O\nX . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place X in row 2, column 2 (cell 5).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 0.3333333333333333, "cell_8": 0.3333333333333333, "cell_9": 0.3333333333333333}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5", "cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XO.XXOO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OOXX.OX..", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.5, "cell_8": 0.29, "cell_9": 0.21}, "solvable": {"false": 0.4, "true": 0.6}, "value": {"0": 0.1, "1": 0.17, "2": 0.73}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:5fba74e84664273c1038f6e0", "state_id": "tic_tac_toe:5fba74e84664273c1038f6e0", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X O\n. . X\nO . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place X in row 3, column 2 (cell 8).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_8": 0.3333333333333333, "cell_4": 0.3333333333333333, "cell_5": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4", "cell_5", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.O..XXXO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXO..XO.X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.22, "cell_5": 0.39999999999999997, "cell_8": 0.38}, "solvable": {"false": 0.49, "true": 0.51}, "value": {"0": 0.15, "1": 0.27, "2": 0.58}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:7ecdf89f91b71372079e2759", "state_id": "grid_navigation:7ecdf89f91b71372079e2759", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A # .\nG . . .\n. . . .\n# # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.5, "west": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 3), (2, 0), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [3, 0], [3, 1], [3, 2]], "position": [0, 1], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.52, "west": 0.48}, "solvable": {"false": 0.12, "true": 0.88}, "value": {"0": 0.73, "1": 0.22, "2": 0.04, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:d5fea5d24436b479fedae089", "state_id": "tic_tac_toe:d5fea5d24436b479fedae089", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . X\nX O O\n. O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.5, "cell_2": 0.5, "cell_9": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.O.OOXX.X:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.XXOO.O.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.19, "cell_7": 0.56, "cell_9": 0.25}, "solvable": {"false": 0.5700000000000001, "true": 0.43}, "value": {"0": 0.27, "1": 0.31, "2": 0.42}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:7455a5fc850b6fd42e02606e", "state_id": "tic_tac_toe:7455a5fc850b6fd42e02606e", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . O\n. . X\nO X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_5": 1.0, "cell_2": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.XX..OXO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.O..XOXO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.29, "cell_4": 0.21, "cell_5": 0.5}, "solvable": {"false": 0.81, "true": 0.19}, "value": {"0": 0.21, "1": 0.46, "2": 0.33}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:f6fadee0bcc7a9fba7a42c80", "state_id": "grid_navigation:f6fadee0bcc7a9fba7a42c80", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nA . . .\n. G . #\n. # # .\n# . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.5, "south": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 3), (1, 2), (2, 2), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 3], [2, 1], [2, 2], [3, 0]], "position": [0, 0], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.65, "south": 0.35}, "solvable": {"false": 0.18999999999999995, "true": 0.81}, "value": {"0": 0.53, "1": 0.35, "2": 0.09, "3": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:92f6d313bea2446a291eae63", "state_id": "grid_navigation:92f6d313bea2446a291eae63", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A # .\n. . . .\n. . . .\n. . # G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 0.0, "south": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (3, 1))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [3, 2]], "position": [0, 1], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.08, "south": 0.92}, "solvable": {"false": 0.09999999999999998, "true": 0.9}, "value": {"0": 0, "1": 0.26, "2": 0.6799999999999999, "3": 0.06}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:6f36f85e13eeb80c17a7e6ca", "state_id": "grid_navigation:6f36f85e13eeb80c17a7e6ca", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. . G #\n. # # .\n. # . A", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (1, 2), (2, 2), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 3], [2, 1], [2, 2], [3, 1]], "position": [3, 3], "goal": [1, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.7, "west": 0.3}, "solvable": {"false": 0.43000000000000005, "true": 0.57}, "value": {"0": 0.02, "1": 0.44, "2": 0.38, "3": 0.16}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:44594724d98700b596c662da", "state_id": "grid_navigation:44594724d98700b596c662da", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # A .\n# . . #\n. G . .\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.0, "south": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 2), (1, 3), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 0], [1, 3]], "position": [0, 2], "goal": [2, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.25, "south": 0.75}, "solvable": {"false": 0.13, "true": 0.87}, "value": {"0": 0.13, "1": 0.74, "2": 0.09, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:406c828bc7160b1fbd48ed7f", "state_id": "grid_navigation:406c828bc7160b1fbd48ed7f", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n. . # .\nG . # .\n. . . A", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((2, 1), (2, 2), (2, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 2], [2, 2]], "position": [3, 3], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.79, "north": 0.21}, "solvable": {"false": 0.37, "true": 0.63}, "value": {"0": 0.02, "1": 0.48, "2": 0.43, "3": 0.07}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:cb443836e7f25d0585393f53", "state_id": "grid_navigation:cb443836e7f25d0585393f53", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . #\n. . G .\n. . . .\n. A . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 0.0, "east": 0.5, "north": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0),)", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3]], "position": [3, 1], "goal": [1, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.86, "east": 0.07, "west": 0.07}, "solvable": {"false": 0.06999999999999995, "true": 0.93}, "value": {"0": 0.17, "1": 0.78, "2": 0.05, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:247fcd73b726bfe3e69392f2", "state_id": "grid_navigation:247fcd73b726bfe3e69392f2", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. . . A\n# . # .\n. G . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 0.5, "south": 0.5, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (1, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[2, 0], [2, 2]], "position": [1, 3], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.02, "west": 0.46, "south": 0.52}, "solvable": {"false": 0.30000000000000004, "true": 0.7}, "value": {"0": 0.04, "1": 0.63, "2": 0.22, "3": 0.11}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:5b2b3a500ac015a02a30d9b2", "state_id": "tic_tac_toe:5b2b3a500ac015a02a30d9b2", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X .\nO O X\n. X X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_7", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 1.0, "cell_3": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOXOXXX.:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OX.OOX.XX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.52, "cell_7": 0.48}, "solvable": {"false": 0.85, "true": 0.15}, "value": {"0": 0.41, "1": 0.37, "2": 0.22}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:b9d680129186a5bef8d8eb80", "state_id": "tic_tac_toe:b9d680129186a5bef8d8eb80", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X .\nX O O\n. O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place O in row 1, column 3 (cell 3).", "cell_7": "Place O in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_3": 0.5, "cell_7": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXXOOXX.:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XX.XOO.OX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.49, "cell_7": 0.51}, "solvable": {"false": 0.83, "true": 0.17}, "value": {"0": 0.4, "1": 0.41, "2": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:86fc5e136fabefca4cc5aceb", "state_id": "grid_navigation:86fc5e136fabefca4cc5aceb", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG . . #\n. . . .\n# # . .\n# # . A", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 2}, "gold_probs": {"action": {"north": 0.5, "west": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (0, 3), (1, 2), (1, 3), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [2, 0], [2, 1], [3, 0], [3, 1]], "position": [3, 3], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.12, "north": 0.88}, "solvable": {"false": 0.24, "true": 0.76}, "value": {"0": 0.01, "1": 0.26, "2": 0.65, "3": 0.08}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:c33ce605b9e6b272cb6192c6", "state_id": "tic_tac_toe:c33ce605b9e6b272cb6192c6", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X .\nO . O\nX . X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place O in row 3, column 2 (cell 8).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_8": 0.0, "cell_5": 1.0, "cell_1": 0.0, "cell_3": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXX...OX:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".X.O.OX.X", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.06, "cell_8": 0.23, "cell_3": 0.05, "cell_5": 0.66}, "solvable": {"false": 0.81, "true": 0.19}, "value": {"0": 0.3, "1": 0.38, "2": 0.32}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:17ebf43bc15ff35b37b597d2", "state_id": "tic_tac_toe:17ebf43bc15ff35b37b597d2", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X O\nX . .\nO X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place X in row 2, column 3 (cell 6).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.0, "cell_1": 0.0, "cell_5": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.XOX..OXO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XOX..OXO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.54, "cell_6": 0.24, "cell_1": 0.22}, "solvable": {"false": 0.75, "true": 0.25}, "value": {"0": 0.27, "1": 0.49, "2": 0.24}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:f6dd15a1ce384babedb46fa0", "state_id": "grid_navigation:f6dd15a1ce384babedb46fa0", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# G # #\n# . . .\n. # A #\n. . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 2), (0, 3), (1, 0), (2, 1), (2, 3), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [0, 3], [1, 0], [2, 1], [2, 3], [3, 3]], "position": [2, 2], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.67, "south": 0.33}, "solvable": {"false": 0.45999999999999996, "true": 0.54}, "value": {"0": 0.03, "1": 0.42, "2": 0.33, "3": 0.22}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:72101ee66158c2aa3cd73a34", "state_id": "grid_navigation:72101ee66158c2aa3cd73a34", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # G .\n# . # .\n# A . .\n. . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.0, "south": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (0, 3), (1, 1), (1, 3), (2, 3), (3, 0))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [1, 0], [1, 2], [2, 0], [3, 3]], "position": [2, 1], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.63, "east": 0.3, "south": 0.07}, "solvable": {"false": 0.4, "true": 0.6}, "value": {"0": 0.03, "1": 0.41, "2": 0.38, "3": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:4988918e911916a26dee2be6", "state_id": "tic_tac_toe:4988918e911916a26dee2be6", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X O\n. X .\n. . O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place O in row 2, column 1 (cell 4).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_8": "Place O in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_6", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_6": 0.5, "cell_7": 0.0, "cell_8": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_6", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O.X.XXO:O", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XXO.X...O", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.43, "cell_6": 0.21, "cell_8": 0.18, "cell_7": 0.18}, "solvable": {"false": 0.91, "true": 0.09}, "value": {"0": 0.49, "1": 0.37, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:e130861f82cc0395f119c788", "state_id": "grid_navigation:e130861f82cc0395f119c788", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . #\nA G . #\n. # . .\n. . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 1.0, "north": 0.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 3), (1, 3), (2, 1), (3, 2))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 3], [2, 1], [3, 2]], "position": [1, 0], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.06, "north": 0.21, "east": 0.73}, "solvable": {"false": 0.040000000000000036, "true": 0.96}, "value": {"0": 0.99, "1": 0, "2": 0, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:dd0475b06b556b805ae58626", "state_id": "tic_tac_toe:dd0475b06b556b805ae58626", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X O\n. . X\nX O O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.0, "cell_4": 1.0, "cell_5": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XX.OOXO:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XO..XXOO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.34, "cell_4": 0.32, "cell_1": 0.34}, "solvable": {"false": 0.69, "true": 0.31}, "value": {"0": 0.33, "1": 0.36, "2": 0.31}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:f3253eac158c748ba177457f", "state_id": "grid_navigation:f3253eac158c748ba177457f", "family_id": "grid_navigation_bfs_v1", "split": "train", "state": "Current game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . #\nG # A .\n. . # .\n. . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 1.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 1), (2, 2), (3, 0), (3, 3))", "template_id": "grid_navigation:train:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 1], [2, 2], [3, 3]], "position": [1, 2], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.59, "east": 0.41}, "solvable": {"false": 0.65, "true": 0.35}, "value": {"0": 0.05, "1": 0.39, "2": 0.29, "3": 0.27}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:f303e2c95859c783d2bded49", "state_id": "tic_tac_toe:f303e2c95859c783d2bded49", "family_id": "tic_tac_toe_minimax_v1", "split": "train", "state": "Current game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X O\n. X .\n. O O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_7": 0.3333333333333333, "cell_4": 0.3333333333333333, "cell_6": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4", "cell_6", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOXXO.O:X", "template_id": "tic_tac_toe:train:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XXO.X..OO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.31, "cell_4": 0.43, "cell_6": 0.26}, "solvable": {"false": 0.5900000000000001, "true": 0.41}, "value": {"0": 0.21, "1": 0.4, "2": 0.39}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:333cf30b0343539c590a9c9a", "state_id": "grid_navigation:333cf30b0343539c590a9c9a", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG . # #\nA . # .\n# . . #\n. . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 1.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (0, 3), (1, 2), (2, 0), (2, 3), (3, 2))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [1, 2], [2, 0], [2, 3], [3, 2]], "position": [1, 0], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.86, "east": 0.14}, "solvable": {"false": 0.31000000000000005, "true": 0.69}, "value": {"0": 0.4, "1": 0.36, "2": 0.13, "3": 0.11}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:e071dea538bde195d47a3d7d", "state_id": "grid_navigation:e071dea538bde195d47a3d7d", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # # .\n. . # G\n# # # A\n. # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (1, 1), (1, 2), (1, 3), (2, 0), (2, 2), (2, 3), (3, 0), (3, 2))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [0, 2], [1, 2], [2, 0], [2, 1], [2, 2], [3, 1], [3, 2]], "position": [2, 3], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.84, "south": 0.16}, "solvable": {"false": 0.55, "true": 0.45}, "value": {"0": 0.22, "1": 0.3, "2": 0.15, "3": 0.33}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:8854fcd843ae09633a1142ac", "state_id": "tic_tac_toe:8854fcd843ae09633a1142ac", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . .\nX . X\n. O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place O in row 1, column 2 (cell 2).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_2": 0.25, "cell_7": 0.25, "cell_5": 0.25, "cell_3": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_3", "cell_5", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OX.XXO.:O", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O..X.X.OX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.19, "cell_7": 0.21, "cell_2": 0.16, "cell_5": 0.44}, "solvable": {"false": 0.9, "true": 0.1}, "value": {"0": 0.54, "1": 0.3, "2": 0.16}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:20a1596920e53d0073f4d79f", "state_id": "grid_navigation:20a1596920e53d0073f4d79f", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # . #\n. . # .\n. . . A\nG # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.0, "west": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 2), (2, 2), (3, 0), (3, 1), (3, 3))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [0, 3], [1, 2], [3, 1], [3, 2]], "position": [2, 3], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.13, "west": 0.7, "south": 0.17}, "solvable": {"false": 0.26, "true": 0.74}, "value": {"0": 0.03, "1": 0.42, "2": 0.38, "3": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:3e31f23d39e8f27ef640d127", "state_id": "tic_tac_toe:3e31f23d39e8f27ef640d127", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . O\nX X O\n. O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_7", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.5, "cell_2": 0.0, "cell_9": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_7", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.O.OXXO.X:X", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.OXXO.O.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.25, "cell_7": 0.46, "cell_9": 0.29}, "solvable": {"false": 0.74, "true": 0.26}, "value": {"0": 0.36, "1": 0.37, "2": 0.27}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:994efb03dbfa8a5b227dc85d", "state_id": "tic_tac_toe:994efb03dbfa8a5b227dc85d", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X O\nO . .\nX X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.3333333333333333, "cell_6": 0.3333333333333333, "cell_5": 0.3333333333333333}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5", "cell_6", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OX.XXOO:X", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXOO..XX.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.55, "cell_6": 0.17, "cell_5": 0.28}, "solvable": {"false": 0.39, "true": 0.61}, "value": {"0": 0.12, "1": 0.15, "2": 0.73}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:706e9f06a03a858739833f77", "state_id": "tic_tac_toe:706e9f06a03a858739833f77", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . O\nX O X\n. X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_7": "Place O in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.25, "cell_2": 0.25, "cell_1": 0.25, "cell_7": 0.25}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_2", "cell_7", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXOX.X.:O", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..OXOX.X.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.22, "cell_1": 0.31, "cell_9": 0.24, "cell_7": 0.23}, "solvable": {"false": 0.91, "true": 0.09}, "value": {"0": 0.3, "1": 0.5, "2": 0.2}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:c97753e5ae4f469f1224843d", "state_id": "tic_tac_toe:c97753e5ae4f469f1224843d", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O .\n. O X\nO X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place X in row 1, column 3 (cell 3).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_3": 1.0, "cell_4": 0.0, "cell_9": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXXO..XO:X", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XO..OXOX.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.28, "cell_9": 0.47, "cell_4": 0.25}, "solvable": {"false": 0.78, "true": 0.22}, "value": {"0": 0.41000000000000003, "1": 0.37, "2": 0.22}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:745e57abfbc8996356b160e5", "state_id": "grid_navigation:745e57abfbc8996356b160e5", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n# . . .\n# G # #\n# A . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 1.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 1), (0, 2), (2, 1), (3, 0), (3, 1))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [2, 0], [2, 2], [2, 3], [3, 0], [3, 3]], "position": [3, 1], "goal": [2, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.66, "east": 0.34}, "solvable": {"false": 0.30000000000000004, "true": 0.7}, "value": {"0": 0.41, "1": 0.38, "2": 0.09, "3": 0.12}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:dc49dae0ae2ccdc8a58c31db", "state_id": "tic_tac_toe:dc49dae0ae2ccdc8a58c31db", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O X\nO . .\n. O X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.0, "cell_5": 0.5, "cell_6": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXO..XOX:X", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOXO...OX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.62, "cell_7": 0.19, "cell_6": 0.19}, "solvable": {"false": 0.74, "true": 0.26}, "value": {"0": 0.18, "1": 0.46, "2": 0.36}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:d896b943317c5edbd43fcc86", "state_id": "grid_navigation:d896b943317c5edbd43fcc86", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . #\n# . A G\n. . . .\n. # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south.", "west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "south": 0.0, "west": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (1, 3), (2, 3), (3, 0), (3, 1))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 3], [1, 0], [3, 1], [3, 2]], "position": [1, 2], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.01, "north": 0, "west": 0.01, "east": 0.98}, "solvable": {"false": 0.020000000000000018, "true": 0.98}, "value": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:ec9807802da066e186543409", "state_id": "tic_tac_toe:ec9807802da066e186543409", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X .\n. O .\nO X X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place X in row 2, column 3 (cell 6).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_3": "Place X in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_6": 0.3333333333333333, "cell_4": 0.3333333333333333, "cell_3": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_4", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XXOXO.O:X", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OX..O.OXX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.38, "cell_3": 0.35, "cell_6": 0.27}, "solvable": {"false": 0.62, "true": 0.38}, "value": {"0": 0.29, "1": 0.31, "2": 0.4}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:3c6d27084e9e53391709956a", "state_id": "tic_tac_toe:3c6d27084e9e53391709956a", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X O\nX . .\nO . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place O in row 3, column 2 (cell 8).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_6": "Place O in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_8": 0.25, "cell_5": 0.25, "cell_9": 0.25, "cell_6": 0.25}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5", "cell_6", "cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O..XOXX:O", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XXOX..O..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.14, "cell_9": 0.1, "cell_5": 0.66, "cell_8": 0.1}, "solvable": {"false": 0.92, "true": 0.08}, "value": {"0": 0.72, "1": 0.18, "2": 0.1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:ad31d2e9953534b3dcd00aee", "state_id": "tic_tac_toe:ad31d2e9953534b3dcd00aee", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X X\nX . O\n. O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_9": 0.3333333333333333, "cell_7": 0.3333333333333333, "cell_5": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5", "cell_7", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.O.O.XXXO:X", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXXX.O.O.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.25, "cell_9": 0.2, "cell_5": 0.55}, "solvable": {"false": 0.63, "true": 0.37}, "value": {"0": 0.26, "1": 0.41, "2": 0.33}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:2fb54d6f7ad94c1fee331f9d", "state_id": "grid_navigation:2fb54d6f7ad94c1fee331f9d", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# G . .\n# . . #\n. . . #\n. A . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.0, "west": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (1, 0), (1, 3), (2, 3), (3, 3))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 0], [1, 3], [2, 3], [3, 3]], "position": [3, 1], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.11, "north": 0.73, "west": 0.16}, "solvable": {"false": 0.22999999999999998, "true": 0.77}, "value": {"0": 0.02, "1": 0.51, "2": 0.32, "3": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:021e638a2f914d9195088a74", "state_id": "grid_navigation:021e638a2f914d9195088a74", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG . . #\n# # . #\n. A # #\n# . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.5, "south": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 3), (1, 1), (2, 2), (2, 3), (3, 0), (3, 1), (3, 2), (3, 3))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 0], [1, 1], [1, 3], [2, 2], [2, 3], [3, 0], [3, 2], [3, 3]], "position": [2, 1], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.38, "west": 0.62}, "solvable": {"false": 0.51, "true": 0.49}, "value": {"0": 0.08, "1": 0.41, "2": 0.23, "3": 0.28}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:8bc9d5ab36f6028775c5dcc6", "state_id": "grid_navigation:8bc9d5ab36f6028775c5dcc6", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # #\n. . . G\n. A # .\n# . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.0, "north": 1.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (1, 0), (1, 2), (3, 0), (3, 3))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [0, 3], [2, 2], [3, 0]], "position": [2, 1], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.2, "north": 0.7, "south": 0.1}, "solvable": {"false": 0.17000000000000004, "true": 0.83}, "value": {"0": 0.13, "1": 0.63, "2": 0.17, "3": 0.07}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:2d53cfc0ce6ddc1006157e94", "state_id": "tic_tac_toe:2d53cfc0ce6ddc1006157e94", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . .\n. X O\n. . O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_8": "Place X in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_2": 0.0, "cell_3": 1.0, "cell_7": 0.0, "cell_8": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O.XOX..:X", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X...XO..O", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.18, "cell_7": 0.17, "cell_2": 0.22, "cell_4": 0.2, "cell_8": 0.23}, "solvable": {"false": 0.84, "true": 0.16}, "value": {"0": 0.34, "1": 0.45, "2": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:92d98a727f6eae7ae56e2077", "state_id": "grid_navigation:92d98a727f6eae7ae56e2077", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # . #\n. # # .\n. . G A\n# . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 1.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 3), (1, 2), (2, 2), (2, 3), (3, 0), (3, 3))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [0, 3], [1, 1], [1, 2], [3, 0], [3, 3]], "position": [2, 3], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.05, "west": 0.95}, "solvable": {"false": 0.08999999999999997, "true": 0.91}, "value": {"0": 0.86, "1": 0.1, "2": 0.03, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:b6fe7b865892738dbbdba5b2", "state_id": "grid_navigation:b6fe7b865892738dbbdba5b2", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . A .\nG . . .\n# # . #\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.0, "south": 0.5, "west": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 2), (3, 2))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[2, 0], [2, 1], [2, 3]], "position": [0, 2], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.01, "south": 0.15, "west": 0.84}, "solvable": {"false": 0.12, "true": 0.88}, "value": {"0": 0.2, "1": 0.64, "2": 0.12, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:97bb2d037cc9fb7bc4504530", "state_id": "tic_tac_toe:97bb2d037cc9fb7bc4504530", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X O\n. . O\nX O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place X in row 2, column 2 (cell 5).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_4": "Place X in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 0.0, "cell_9": 0.5, "cell_4": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOO.XX.X:X", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XXO..OXO.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.32, "cell_9": 0.47, "cell_4": 0.21}, "solvable": {"false": 0.76, "true": 0.24}, "value": {"0": 0.3, "1": 0.45, "2": 0.25}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:1c6a677bbaef40e852f0585c", "state_id": "tic_tac_toe:1c6a677bbaef40e852f0585c", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X O\nO O .\n. X X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place O in row 2, column 3 (cell 6).", "cell_7": "Place O in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_6", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.5, "cell_7": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_6", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXXOXX.O:O", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XXOOO..XX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.67, "cell_7": 0.33}, "solvable": {"false": 0.79, "true": 0.21}, "value": {"0": 0.3, "1": 0.36, "2": 0.34}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:afb4ff0f839aa62a75fad20a", "state_id": "grid_navigation:afb4ff0f839aa62a75fad20a", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\n. . G .\nA # . .\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 2), (2, 2))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [2, 1]], "position": [2, 0], "goal": [1, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.25, "north": 0.75}, "solvable": {"false": 0.16000000000000003, "true": 0.84}, "value": {"0": 0.1, "1": 0.71, "2": 0.15, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:3b2cc34bf80930183b327743", "state_id": "tic_tac_toe:3b2cc34bf80930183b327743", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . O\nX X .\n. . X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place O in row 3, column 2 (cell 8).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_6": "Place O in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_8": 0.0, "cell_2": 1.0, "cell_7": 0.0, "cell_6": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XXX.O.O:O", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.OXX...X", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.17, "cell_7": 0.15, "cell_2": 0.16, "cell_6": 0.52}, "solvable": {"false": 0.86, "true": 0.14}, "value": {"0": 0.47, "1": 0.36, "2": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:88b67d80d2e01653a7d07a05", "state_id": "grid_navigation:88b67d80d2e01653a7d07a05", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . A\n. . . .\nG . # #\n# . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 0.5, "south": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 3), (2, 0), (2, 1), (3, 1))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [2, 2], [2, 3], [3, 0], [3, 2]], "position": [0, 3], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.51, "west": 0.49}, "solvable": {"false": 0.15000000000000002, "true": 0.85}, "value": {"0": 0.01, "1": 0.63, "2": 0.29, "3": 0.07}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:27105926830f0bc7bbcd34cf", "state_id": "grid_navigation:27105926830f0bc7bbcd34cf", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # .\n# . A .\n. # # G\n. . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 1.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (1, 1), (1, 3), (2, 1), (2, 3), (3, 2))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [1, 0], [2, 1], [2, 2], [3, 3]], "position": [1, 2], "goal": [2, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.66, "west": 0.34}, "solvable": {"false": 0.55, "true": 0.45}, "value": {"0": 0.12, "1": 0.43, "2": 0.22, "3": 0.23}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:909e1e80331eb17724cb5f45", "state_id": "tic_tac_toe:909e1e80331eb17724cb5f45", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X .\n. X O\n. . O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.0, "cell_3": 0.5, "cell_4": 0.0, "cell_8": 0.5, "cell_7": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....XXOO.:X", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".X..XO..O", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.21, "cell_3": 0.13, "cell_7": 0.26, "cell_8": 0.17, "cell_1": 0.23}, "solvable": {"false": 0.73, "true": 0.27}, "value": {"0": 0.35, "1": 0.35, "2": 0.3}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:bad5dc473dab4b80017c23d8", "state_id": "grid_navigation:bad5dc473dab4b80017c23d8", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # .\n# . . #\n# . A .\n. G . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east.", "north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.5, "east": 0.0, "north": 0.0, "west": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (2, 0), (2, 3), (3, 0), (3, 2))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [1, 0], [1, 3], [2, 0]], "position": [2, 2], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.51, "north": 0.04, "south": 0.44, "east": 0.01}, "solvable": {"false": 0.12, "true": 0.88}, "value": {"0": 0.25, "1": 0.63, "2": 0.07, "3": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:09a9fcbe6420de93d556f1f6", "state_id": "grid_navigation:09a9fcbe6420de93d556f1f6", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # G\n. # . .\n# A . .\n# . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 2), (1, 2), (2, 3), (3, 1), (3, 3))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [1, 1], [2, 0], [3, 0], [3, 2]], "position": [2, 1], "goal": [0, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.91, "south": 0.09}, "solvable": {"false": 0.5700000000000001, "true": 0.43}, "value": {"0": 0.02, "1": 0.33, "2": 0.35, "3": 0.3}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:b73cc3d3225e9d7b6715b9cf", "state_id": "tic_tac_toe:b73cc3d3225e9d7b6715b9cf", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X .\nX O X\n. O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place O in row 1, column 3 (cell 3).", "cell_7": "Place O in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_3": 1.0, "cell_7": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXXOXOX.:O", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OX.XOX.OX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.6, "cell_3": 0.4}, "solvable": {"false": 0.9299999999999999, "true": 0.07}, "value": {"0": 0.71, "1": 0.23, "2": 0.06}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:6633607d723d5edfabde6c66", "state_id": "tic_tac_toe:6633607d723d5edfabde6c66", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X X\n. O X\nO X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place O in row 2, column 1 (cell 4).", "cell_1": "Place O in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_1": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXOXXXO:O", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XX.OXOXO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.6, "cell_4": 0.4}, "solvable": {"false": 0.88, "true": 0.12}, "value": {"0": 0.44, "1": 0.39, "2": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:02e96a9fe5e28dc9b13bee57", "state_id": "tic_tac_toe:02e96a9fe5e28dc9b13bee57", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\n. O O\n. X X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place X in row 1, column 3 (cell 3).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_1": "Place X in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_7", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_3": 0.0, "cell_4": 0.0, "cell_7": 1.0, "cell_2": 0.0, "cell_1": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....OO.XX:X", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "....OO.XX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.27, "cell_7": 0.35, "cell_3": 0.1, "cell_1": 0.15, "cell_2": 0.13}, "solvable": {"false": 0.6, "true": 0.4}, "value": {"0": 0.22, "1": 0.34, "2": 0.44}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:543f4ceda4e1c72d246f72bf", "state_id": "tic_tac_toe:543f4ceda4e1c72d246f72bf", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO O .\nO . X\nX . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place X in row 2, column 2 (cell 5).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_3": "Place X in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 0.0, "cell_8": 0.5, "cell_3": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOX.OX.X:X", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OO.O.XX.X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.2, "cell_8": 0.23, "cell_5": 0.56}, "solvable": {"false": 0.6, "true": 0.4}, "value": {"0": 0.4, "1": 0.27, "2": 0.33}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:ca0b4bb94407809644e92dee", "state_id": "tic_tac_toe:ca0b4bb94407809644e92dee", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . O\nX O X\n. O X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_7": 0.3333333333333333, "cell_1": 0.3333333333333333, "cell_2": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_2", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXOX.OX:X", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..OXOX.OX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.4, "cell_2": 0.2, "cell_7": 0.39999999999999997}, "solvable": {"false": 0.6799999999999999, "true": 0.32}, "value": {"0": 0.32, "1": 0.4, "2": 0.28}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:593ae89c1f152814ea0c5c3b", "state_id": "grid_navigation:593ae89c1f152814ea0c5c3b", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . A\n. . . .\n. . # G\n# . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 1.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (1, 1), (3, 0))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[2, 2], [3, 0], [3, 3]], "position": [0, 3], "goal": [2, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.19, "south": 0.81}, "solvable": {"false": 0.18000000000000005, "true": 0.82}, "value": {"0": 0.01, "1": 0.6, "2": 0.29, "3": 0.1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:69b680f0d4f4f8387ac233a7", "state_id": "grid_navigation:69b680f0d4f4f8387ac233a7", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . A .\n# . . G\n. . . .\n# . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "east": 0.5, "south": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((3, 0), (3, 1), (3, 3))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 0], [3, 0]], "position": [0, 2], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.21, "west": 0.05, "east": 0.74}, "solvable": {"false": 0.06999999999999995, "true": 0.93}, "value": {"0": 0.66, "1": 0.3, "2": 0.03, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:eb45f5d4d02ec420efbaa956", "state_id": "grid_navigation:eb45f5d4d02ec420efbaa956", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A . .\n. # . G\n. # . .\n. . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 3), (2, 1), (2, 2))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 1], [2, 1], [3, 2]], "position": [0, 1], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.1, "east": 0.9}, "solvable": {"false": 0.25, "true": 0.75}, "value": {"0": 0.06, "1": 0.71, "2": 0.17, "3": 0.06}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:56719bae51ddad611d9f4ac1", "state_id": "grid_navigation:56719bae51ddad611d9f4ac1", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # #\n. . # .\n. A # #\n. # G .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 0), (1, 1), (2, 1), (3, 0), (3, 1))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [1, 2], [2, 2], [2, 3], [3, 1]], "position": [2, 1], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.47, "north": 0.53}, "solvable": {"false": 0.63, "true": 0.37}, "value": {"0": 0.15, "1": 0.35000000000000003, "2": 0.2, "3": 0.3}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:8b4d5c9b8bdbd7d9e2d1af5a", "state_id": "tic_tac_toe:8b4d5c9b8bdbd7d9e2d1af5a", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X O\n. . .\nX . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place X in row 2, column 3 (cell 6).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_8", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.0, "cell_4": 0.0, "cell_8": 1.0, "cell_9": 0.0, "cell_5": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O..XX.O:X", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXO...X..", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.51, "cell_9": 0.09, "cell_8": 0.09, "cell_6": 0.06, "cell_4": 0.25}, "solvable": {"false": 0.84, "true": 0.16}, "value": {"0": 0.26, "1": 0.51, "2": 0.23}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:875daa108051d1ccf3766987", "state_id": "tic_tac_toe:875daa108051d1ccf3766987", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . .\nX O X\n. . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place X in row 1, column 3 (cell 3).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_3": 0.2, "cell_9": 0.2, "cell_2": 0.2, "cell_8": 0.2, "cell_7": 0.2}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_3", "cell_7", "cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...XOX..O:X", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O..XOX...", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.15, "cell_7": 0.24, "cell_3": 0.21, "cell_8": 0.15, "cell_2": 0.25}, "solvable": {"false": 0.78, "true": 0.22}, "value": {"0": 0.2, "1": 0.49, "2": 0.31}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:fc28fb280580938d827383db", "state_id": "grid_navigation:fc28fb280580938d827383db", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # .\nG # # #\n# # A .\n# # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "south": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (0, 3), (1, 0), (1, 1), (1, 2), (1, 3), (2, 0), (2, 1), (3, 1))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [1, 1], [1, 2], [1, 3], [2, 0], [2, 1], [3, 0], [3, 1]], "position": [2, 2], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.63, "south": 0.37}, "solvable": {"false": 0.78, "true": 0.22}, "value": {"0": 0.03, "1": 0.14, "2": 0.15, "3": 0.6799999999999999}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:b3ca27f3f0695f7d8ac8989c", "state_id": "tic_tac_toe:b3ca27f3f0695f7d8ac8989c", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X .\nX X O\n. O O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.5, "cell_3": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOXXOXX.:O", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XX.XXO.OO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.55, "cell_7": 0.45}, "solvable": {"false": 0.85, "true": 0.15}, "value": {"0": 0.44, "1": 0.31, "2": 0.25}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:6429a0a5f4beacbb44d3dc3b", "state_id": "tic_tac_toe:6429a0a5f4beacbb44d3dc3b", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . .\nX . O\nO X X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place X in row 1, column 3 (cell 3).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_3": 0.3333333333333333, "cell_5": 0.3333333333333333, "cell_2": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_3", "cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OO.XXXO:X", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O..X.OOXX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.26, "cell_5": 0.46, "cell_2": 0.28}, "solvable": {"false": 0.6699999999999999, "true": 0.33}, "value": {"0": 0.23, "1": 0.34, "2": 0.43}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:471c0ab381833e92c06f78e5", "state_id": "grid_navigation:471c0ab381833e92c06f78e5", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n. # # #\n. A G #\n. # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (0, 2), (1, 2), (1, 3), (2, 0), (2, 2))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 1], [1, 2], [1, 3], [2, 3], [3, 1]], "position": [2, 1], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.98, "west": 0.02}, "solvable": {"false": 0.06999999999999995, "true": 0.93}, "value": {"0": 0.96, "1": 0.01, "2": 0.01, "3": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:fccbc0876bbdda40191c3593", "state_id": "grid_navigation:fccbc0876bbdda40191c3593", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # .\n. A # .\n. # . .\n# . G #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 3), (1, 2), (2, 1), (3, 1), (3, 3))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [1, 2], [2, 1], [3, 0], [3, 3]], "position": [1, 1], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.28, "west": 0.72}, "solvable": {"false": 0.39, "true": 0.61}, "value": {"0": 0.03, "1": 0.47, "2": 0.33, "3": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:526a0f8017a1f55fe31f3ee7", "state_id": "tic_tac_toe:526a0f8017a1f55fe31f3ee7", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X X\n. O O\nO X X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_4": "Place O in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.0, "cell_4": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXOXXOX:O", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XX.OOOXX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.5, "cell_1": 0.5}, "solvable": {"false": 0.8, "true": 0.2}, "value": {"0": 0.48, "1": 0.27, "2": 0.25}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:f3788570fc62728f3d92a8eb", "state_id": "tic_tac_toe:f3788570fc62728f3d92a8eb", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X .\nX O O\n. O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_3": "Place X in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.3333333333333333, "cell_9": 0.3333333333333333, "cell_3": 0.3333333333333333}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_7", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.O.OOX.XX:X", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XX.XOO.O.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.34, "cell_3": 0.42, "cell_9": 0.24}, "solvable": {"false": 0.73, "true": 0.27}, "value": {"0": 0.32, "1": 0.29, "2": 0.39}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:0d0c9584582e57ea6217c54e", "state_id": "grid_navigation:0d0c9584582e57ea6217c54e", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # #\n. . . .\nA . . #\n. . # G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.3333333333333333, "south": 0.3333333333333333, "north": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 3), (1, 0), (1, 3), (3, 3))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [0, 3], [2, 3], [3, 2]], "position": [2, 0], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.55, "south": 0.34, "north": 0.11}, "solvable": {"false": 0.19999999999999996, "true": 0.8}, "value": {"0": 0.01, "1": 0.47, "2": 0.38, "3": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:8661e21e32acae2776712dd5", "state_id": "tic_tac_toe:8661e21e32acae2776712dd5", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O X\nX . .\nX O .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_6": "Place O in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.0, "cell_5": 1.0, "cell_1": 0.0, "cell_6": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XO.OXX.:O", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".OXX..XO.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.13, "cell_6": 0.19, "cell_9": 0.26, "cell_5": 0.42}, "solvable": {"false": 0.9, "true": 0.1}, "value": {"0": 0.64, "1": 0.22, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:2913f781bd43e05f4fd10343", "state_id": "grid_navigation:2913f781bd43e05f4fd10343", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# G . .\n# . . #\n. . # .\n# # . A", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (1, 0), (1, 3), (2, 2), (3, 0), (3, 1))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 0], [1, 3], [2, 2], [3, 0], [3, 1]], "position": [3, 3], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.8, "west": 0.2}, "solvable": {"false": 0.32999999999999996, "true": 0.67}, "value": {"0": 0.01, "1": 0.29, "2": 0.53, "3": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:cba0cd7995a712593efab7b0", "state_id": "grid_navigation:cba0cd7995a712593efab7b0", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A . #\n# . # .\n# # . G\n. . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"south": 0.3333333333333333, "west": 0.3333333333333333, "east": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 0), (1, 2), (2, 1), (3, 1), (3, 2))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 0], [1, 2], [2, 0], [2, 1], [3, 2]], "position": [0, 1], "goal": [2, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.05, "east": 0.05, "south": 0.9}, "solvable": {"false": 0.33999999999999997, "true": 0.66}, "value": {"0": 0.01, "1": 0.45, "2": 0.34, "3": 0.2}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:64d8b75bfb3b9ff363b40131", "state_id": "tic_tac_toe:64d8b75bfb3b9ff363b40131", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . .\nO O X\nX O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place O in row 1, column 2 (cell 2).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 1.0, "cell_3": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XXOOXOX:O", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X..OOXXOX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.43, "cell_2": 0.57}, "solvable": {"false": 0.89, "true": 0.11}, "value": {"0": 0.37, "1": 0.46, "2": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:e5edc91280a1a86ebfd3c01b", "state_id": "tic_tac_toe:e5edc91280a1a86ebfd3c01b", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . X\n. . O\nX O .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_9": 0.25, "cell_4": 0.25, "cell_2": 0.25, "cell_5": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_4", "cell_5", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXO..X.X:O", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.X..OXO.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.31, "cell_4": 0.18, "cell_5": 0.3, "cell_2": 0.21}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.5, "1": 0.32, "2": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:012cf3fc89c0d31e088c12af", "state_id": "tic_tac_toe:012cf3fc89c0d31e088c12af", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O O\n. . X\nO X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.0, "cell_4": 0.0, "cell_5": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.XOX..OOX:X", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOO..XOX.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.38, "cell_4": 0.3, "cell_5": 0.32}, "solvable": {"false": 0.78, "true": 0.22}, "value": {"0": 0.26, "1": 0.48, "2": 0.26}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:4c13c27e92362d796f09d7c9", "state_id": "tic_tac_toe:4c13c27e92362d796f09d7c9", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . X\nO O X\n. X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_7": 0.0, "cell_2": 1.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXXO.OXX:O", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.XOOX.XO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.41, "cell_7": 0.59}, "solvable": {"false": 0.88, "true": 0.12}, "value": {"0": 0.54, "1": 0.32, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:cf03de0b99c364d26015347d", "state_id": "grid_navigation:cf03de0b99c364d26015347d", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# G . .\n# . . #\n. A . .\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west.", "east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "west": 0.0, "east": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (1, 0), (1, 3))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 0], [1, 3]], "position": [2, 1], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.14, "north": 0.68, "south": 0.03, "west": 0.15}, "solvable": {"false": 0.21999999999999997, "true": 0.78}, "value": {"0": 0.06, "1": 0.53, "2": 0.26, "3": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:0f364fd511f2e93666bab70b", "state_id": "grid_navigation:0f364fd511f2e93666bab70b", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A . G\n# # # #\n. . . .\n. . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (1, 1), (1, 3), (2, 1), (3, 1))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [1, 1], [1, 2], [1, 3], [3, 2]], "position": [0, 1], "goal": [0, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.89, "west": 0.11}, "solvable": {"false": 0.6799999999999999, "true": 0.32}, "value": {"0": 0.82, "1": 0.05, "2": 0.04, "3": 0.09}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:e82e5052a9041244a4c3a150", "state_id": "tic_tac_toe:e82e5052a9041244a4c3a150", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X .\nX O .\nO . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place O in row 2, column 3 (cell 6).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.0, "cell_9": 0.0, "cell_8": 0.0, "cell_3": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....OXOXX:O", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XX.XO.O..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.28, "cell_8": 0.26, "cell_9": 0.17, "cell_3": 0.29}, "solvable": {"false": 0.88, "true": 0.12}, "value": {"0": 0.6, "1": 0.2, "2": 0.2}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:c72288a20081a91c10dbfaa6", "state_id": "grid_navigation:c72288a20081a91c10dbfaa6", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . #\n. . . #\nG # A .\n. # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 1.0, "south": 0.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (1, 1), (3, 2), (3, 3))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 3], [2, 1], [3, 1]], "position": [2, 2], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.51, "east": 0.15, "south": 0.34}, "solvable": {"false": 0.6599999999999999, "true": 0.34}, "value": {"0": 0.05, "1": 0.34, "2": 0.33, "3": 0.28}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:525c14031ab8794a1c7a67e3", "state_id": "grid_navigation:525c14031ab8794a1c7a67e3", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A # #\n# . . .\n. # # .\n. G # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": false, "value": 3}, "gold_probs": {"action": {"south": 0.5, "west": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 1), (1, 2), (2, 0), (3, 2), (3, 3))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [1, 0], [2, 1], [2, 2], [3, 2]], "position": [0, 1], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.51, "west": 0.49}, "solvable": {"false": 0.48, "true": 0.52}, "value": {"0": 0.04, "1": 0.43, "2": 0.28, "3": 0.25}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:8f42d959b143ecb4f82766f8", "state_id": "grid_navigation:8f42d959b143ecb4f82766f8", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . #\n. . # .\nG A . .\n. # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 1.0, "east": 0.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (2, 2), (3, 3))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 2], [3, 1]], "position": [2, 1], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.97, "north": 0.02, "east": 0.01}, "solvable": {"false": 0.18999999999999995, "true": 0.81}, "value": {"0": 0.81, "1": 0.11, "2": 0.04, "3": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:f76f45dadc5536a5d69ab114", "state_id": "tic_tac_toe:f76f45dadc5536a5d69ab114", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . .\nO O X\n. X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place X in row 1, column 3 (cell 3).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_3": 0.3333333333333333, "cell_7": 0.3333333333333333, "cell_2": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_3", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XXOOOX.:X", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X..OOX.XO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.29, "cell_7": 0.48, "cell_3": 0.23}, "solvable": {"false": 0.6799999999999999, "true": 0.32}, "value": {"0": 0.21, "1": 0.36, "2": 0.43}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:67641381b96f984775e01e3e", "state_id": "tic_tac_toe:67641381b96f984775e01e3e", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O .\nO X .\n. . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place X in row 2, column 3 (cell 6).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_1": "Place X in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.2, "cell_8": 0.2, "cell_7": 0.2, "cell_3": 0.2, "cell_1": 0.2}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_3", "cell_6", "cell_7", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOX..O.:X", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".O.OX...X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.2, "cell_6": 0.27, "cell_8": 0.17, "cell_3": 0.21, "cell_7": 0.15}, "solvable": {"false": 0.75, "true": 0.25}, "value": {"0": 0.19, "1": 0.42, "2": 0.39}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:b2093370e93a61874d25fde5", "state_id": "grid_navigation:b2093370e93a61874d25fde5", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\nG # . .\n# # A .\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.3333333333333333, "south": 0.3333333333333333, "north": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 0), (1, 1), (1, 2))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 1], [2, 0], [2, 1]], "position": [2, 2], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.32, "south": 0.42, "east": 0.26}, "solvable": {"false": 0.5700000000000001, "true": 0.43}, "value": {"0": 0.02, "1": 0.37, "2": 0.41, "3": 0.2}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:353807ed124fac49898cf0d8", "state_id": "tic_tac_toe:353807ed124fac49898cf0d8", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X O\nO X O\nX . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_1": "Place X in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_8", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.5, "cell_8": 0.5, "cell_1": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOXOOX.:X", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XOOXOX..", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.23, "cell_9": 0.25, "cell_8": 0.52}, "solvable": {"false": 0.64, "true": 0.36}, "value": {"0": 0.16, "1": 0.35, "2": 0.49}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:017df13168256e8b9599c559", "state_id": "tic_tac_toe:017df13168256e8b9599c559", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X .\nX . O\n. O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_7": "Place O in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_1": 1.0, "cell_5": 0.0, "cell_3": 0.0, "cell_7": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXX.O.X.:O", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".X.X.O.OX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.15, "cell_7": 0.13, "cell_3": 0.11, "cell_5": 0.61}, "solvable": {"false": 0.92, "true": 0.08}, "value": {"0": 0.34, "1": 0.52, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:5c32e7a71d328f033d0a834c", "state_id": "grid_navigation:5c32e7a71d328f033d0a834c", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . G A\n# . # .\n. . . .\n. . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 1), (1, 3), (3, 0), (3, 1))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [1, 2], [3, 2], [3, 3]], "position": [0, 3], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.91, "south": 0.09}, "solvable": {"false": 0.16000000000000003, "true": 0.84}, "value": {"0": 0.69, "1": 0.2, "2": 0.06, "3": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:cc79ed92bc435fe3230b484b", "state_id": "tic_tac_toe:cc79ed92bc435fe3230b484b", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X O\n. O .\n. X X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_7", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_7": 1.0, "cell_6": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXOXX.O:X", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXO.O..XX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.35, "cell_6": 0.25, "cell_4": 0.4}, "solvable": {"false": 0.6599999999999999, "true": 0.34}, "value": {"0": 0.29, "1": 0.36, "2": 0.35}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:90e1a0dbb0d48adc08156520", "state_id": "grid_navigation:90e1a0dbb0d48adc08156520", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\nG # # #\n# A . .\n. . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"south": 0.5, "east": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 1), (2, 1), (2, 3), (3, 1), (3, 3))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 1], [1, 2], [1, 3], [2, 0], [3, 2], [3, 3]], "position": [2, 1], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.66, "south": 0.34}, "solvable": {"false": 0.62, "true": 0.38}, "value": {"0": 0.09, "1": 0.29, "2": 0.16, "3": 0.46}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:add0d530e3f912b118be57cf", "state_id": "grid_navigation:add0d530e3f912b118be57cf", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # # #\n. . . .\n. . # #\n. A G .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "east": 1.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (1, 1), (3, 0), (3, 1), (3, 2), (3, 3))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [0, 2], [0, 3], [2, 2], [2, 3]], "position": [3, 1], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.02, "east": 0.98, "west": 0}, "solvable": {"false": 0.040000000000000036, "true": 0.96}, "value": {"0": 0.99, "1": 0.01, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:397c35b9d7fa193644643678", "state_id": "tic_tac_toe:397c35b9d7fa193644643678", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO O .\n. . .\nX . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place X in row 2, column 2 (cell 5).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_3": "Place X in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 0.0, "cell_4": 0.0, "cell_8": 0.5, "cell_6": 0.0, "cell_3": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XO..O.X:X", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OO....X.X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.23, "cell_5": 0.32, "cell_4": 0.11, "cell_3": 0.22, "cell_6": 0.12}, "solvable": {"false": 0.62, "true": 0.38}, "value": {"0": 0.18, "1": 0.37, "2": 0.45}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:9f39743adbf0140e5fee4365", "state_id": "grid_navigation:9f39743adbf0140e5fee4365", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # . #\n. . . #\nA G # .\n. . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 1.0, "south": 0.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 2), (1, 2), (2, 3), (3, 0), (3, 1), (3, 3))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [0, 3], [1, 3], [2, 2], [3, 2]], "position": [2, 0], "goal": [2, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.8, "north": 0.15, "south": 0.05}, "solvable": {"false": 0.040000000000000036, "true": 0.96}, "value": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:8c6fc2d08425c92da5b23366", "state_id": "tic_tac_toe:8c6fc2d08425c92da5b23366", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . O\n. X .\nO . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_8": "Place X in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.2, "cell_1": 0.2, "cell_2": 0.2, "cell_6": 0.2, "cell_8": 0.2}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_2", "cell_4", "cell_6", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O.X.O.X:X", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..O.X.O.X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.29, "cell_6": 0.16, "cell_1": 0.2, "cell_2": 0.17, "cell_4": 0.18}, "solvable": {"false": 0.83, "true": 0.17}, "value": {"0": 0.21, "1": 0.51, "2": 0.28}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:79d4b237b630995118426d83", "state_id": "grid_navigation:79d4b237b630995118426d83", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # .\nG # A .\n. . . #\n. . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.0, "south": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (1, 1), (2, 0), (3, 2), (3, 3))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [1, 1], [2, 3], [3, 3]], "position": [1, 2], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.83, "east": 0.17}, "solvable": {"false": 0.54, "true": 0.46}, "value": {"0": 0.11, "1": 0.52, "2": 0.21, "3": 0.16}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:072d7982809b97243c3740bb", "state_id": "grid_navigation:072d7982809b97243c3740bb", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n# . . #\n. . . A\nG . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (2, 0), (2, 3))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [1, 3], [3, 2]], "position": [2, 3], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.27, "west": 0.73}, "solvable": {"false": 0.24, "true": 0.76}, "value": {"0": 0.07, "1": 0.61, "2": 0.23, "3": 0.09}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:5d1b8a9f3cf84ce59aff6c73", "state_id": "grid_navigation:5d1b8a9f3cf84ce59aff6c73", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A G #\n. # . .\n. # . #\n. . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (1, 2), (2, 0), (2, 2), (3, 0))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 1], [2, 1], [2, 3], [3, 3]], "position": [0, 1], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.98, "west": 0.02}, "solvable": {"false": 0.06000000000000005, "true": 0.94}, "value": {"0": 0.97, "1": 0.01, "2": 0.01, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:a90bc5ed8fecdd14fada7c1a", "state_id": "grid_navigation:a90bc5ed8fecdd14fada7c1a", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A . #\nG # . #\n# # # #\n. . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 1.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 1), (1, 2), (2, 2), (2, 3), (3, 0), (3, 1), (3, 2))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 1], [1, 3], [2, 0], [2, 1], [2, 2], [2, 3], [3, 2]], "position": [0, 1], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.13, "west": 0.87}, "solvable": {"false": 0.6599999999999999, "true": 0.34}, "value": {"0": 0.16, "1": 0.14, "2": 0.06, "3": 0.64}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:b2e537e7789ee6193e66686d", "state_id": "grid_navigation:b2e537e7789ee6193e66686d", "family_id": "grid_navigation_bfs_v1", "split": "dev", "state": "Review this game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\n# G . #\nA . . #\n. . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (1, 0), (1, 3), (2, 3), (3, 2))", "template_id": "grid_navigation:dev:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 0], [1, 3], [2, 3], [3, 2]], "position": [2, 0], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.11, "east": 0.89}, "solvable": {"false": 0.19999999999999996, "true": 0.8}, "value": {"0": 0.22, "1": 0.59, "2": 0.1, "3": 0.09}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:38065352b535aaa47e7fc2eb", "state_id": "tic_tac_toe:38065352b535aaa47e7fc2eb", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\n. X O\n. O X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.3333333333333333, "cell_4": 0.0, "cell_3": 0.3333333333333333, "cell_7": 0.3333333333333333, "cell_2": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_3", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....XO.OX:X", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "....XO.OX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.28, "cell_7": 0.21, "cell_3": 0.19, "cell_1": 0.18, "cell_2": 0.14}, "solvable": {"false": 0.83, "true": 0.17}, "value": {"0": 0.26, "1": 0.48, "2": 0.26}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:84cc3cb3edd4ac68aef899ba", "state_id": "tic_tac_toe:84cc3cb3edd4ac68aef899ba", "family_id": "tic_tac_toe_minimax_v1", "split": "dev", "state": "Review this game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . O\nO O X\nX X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place X in row 1, column 2 (cell 2).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.0, "cell_1": 0.0, "cell_9": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OOOXXX.:X", "template_id": "tic_tac_toe:dev:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..OOOXXX.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.76, "cell_2": 0.1, "cell_1": 0.14}, "solvable": {"false": 0.35, "true": 0.65}, "value": {"0": 0.11, "1": 0.2, "2": 0.69}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:81bbf883581d4f8305589a7f", "state_id": "grid_navigation:81bbf883581d4f8305589a7f", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # #\n. . . .\n# . . .\nG A . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.0, "north": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (2, 3), (3, 0), (3, 3))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [2, 0], [3, 3]], "position": [3, 1], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.03, "west": 0.9, "north": 0.07}, "solvable": {"false": 0.17000000000000004, "true": 0.83}, "value": {"0": 0.77, "1": 0.17, "2": 0.04, "3": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:198483f8636fd797b7587f88", "state_id": "grid_navigation:198483f8636fd797b7587f88", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # A .\n. # . .\n. . G #\n. . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.0, "south": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (1, 0), (2, 2), (2, 3))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 1], [2, 3], [3, 2]], "position": [0, 2], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.47, "east": 0.53}, "solvable": {"false": 0.31000000000000005, "true": 0.69}, "value": {"0": 0.06, "1": 0.67, "2": 0.17, "3": 0.1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:bb55ff6243e7d584b11b0fe9", "state_id": "tic_tac_toe:bb55ff6243e7d584b11b0fe9", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . O\nO O X\nX X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place O in row 1, column 2 (cell 2).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_2": 0.0, "cell_9": 1.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.XOXO.XOX:O", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.OOOXXX.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.27, "cell_9": 0.73}, "solvable": {"false": 0.78, "true": 0.22}, "value": {"0": 0.44, "1": 0.31, "2": 0.25}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:54042c1017211b5c0490cc29", "state_id": "grid_navigation:54042c1017211b5c0490cc29", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # #\n. A # .\n# # . #\n# G . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 3), (1, 0), (1, 1), (1, 3), (2, 2), (3, 0), (3, 2), (3, 3))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [0, 3], [1, 2], [2, 0], [2, 1], [2, 3], [3, 0], [3, 3]], "position": [1, 1], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.18, "west": 0.82}, "solvable": {"false": 0.35, "true": 0.65}, "value": {"0": 0.09, "1": 0.61, "2": 0.13, "3": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:917bc81f7ee11655350efa87", "state_id": "tic_tac_toe:917bc81f7ee11655350efa87", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . X\n. X X\nO O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.5, "cell_4": 0.5, "cell_2": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOXX.X.O:X", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.X.XXOO.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.55, "cell_2": 0.11, "cell_4": 0.34}, "solvable": {"false": 0.38, "true": 0.62}, "value": {"0": 0.13, "1": 0.18, "2": 0.69}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:9a4a868ea9a0755d481d1f75", "state_id": "tic_tac_toe:9a4a868ea9a0755d481d1f75", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . .\nX . O\nX X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place X in row 1, column 2 (cell 2).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_2": 0.3333333333333333, "cell_3": 0.3333333333333333, "cell_5": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_3", "cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OO.XOXX:X", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O..X.OXXO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.33, "cell_5": 0.45, "cell_3": 0.22}, "solvable": {"false": 0.53, "true": 0.47}, "value": {"0": 0.21, "1": 0.21, "2": 0.58}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:ee4c062249754c2e3788773a", "state_id": "tic_tac_toe:ee4c062249754c2e3788773a", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O .\nX . X\n. O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place X in row 1, column 3 (cell 3).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_3": 0.0, "cell_5": 1.0, "cell_1": 0.0, "cell_7": 0.0, "cell_9": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.O.X.X.O.:X", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".O.X.X.O.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.06, "cell_5": 0.7, "cell_3": 0.06, "cell_9": 0.05, "cell_1": 0.13}, "solvable": {"false": 0.77, "true": 0.23}, "value": {"0": 0.2, "1": 0.53, "2": 0.27}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:b276bb54b396fa4fb8343eef", "state_id": "grid_navigation:b276bb54b396fa4fb8343eef", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # G\n. # . .\n. A # .\n. . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.5, "south": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (1, 2), (2, 1), (3, 0))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 1], [2, 2], [3, 3]], "position": [2, 1], "goal": [0, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.61, "south": 0.39}, "solvable": {"false": 0.48, "true": 0.52}, "value": {"0": 0.02, "1": 0.45, "2": 0.34, "3": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:0e33ba0f5e78c45d2e992f19", "state_id": "grid_navigation:0e33ba0f5e78c45d2e992f19", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # # #\n# . . .\n. . # #\nA G # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 2), (1, 3), (2, 0), (3, 0), (3, 1), (3, 2), (3, 3))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [0, 2], [0, 3], [1, 0], [2, 2], [2, 3], [3, 2]], "position": [3, 0], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.46, "east": 0.54}, "solvable": {"false": 0.06000000000000005, "true": 0.94}, "value": {"0": 0.95, "1": 0.03, "2": 0.01, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:f6fa69bdf9cf0fd39880949d", "state_id": "tic_tac_toe:f6fa69bdf9cf0fd39880949d", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X X\nX . O\n. O O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_1": "Place X in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.5, "cell_5": 0.0, "cell_1": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOX.O.XX:X", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XXX.O.OO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.15, "cell_5": 0.59, "cell_7": 0.26}, "solvable": {"false": 0.56, "true": 0.44}, "value": {"0": 0.36, "1": 0.31, "2": 0.33}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:f763daea5f82c1c324b24370", "state_id": "grid_navigation:f763daea5f82c1c324b24370", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # .\n. # # A\n. . # G\n# . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 1.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (1, 1), (1, 2), (1, 3), (2, 2), (2, 3), (3, 0))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [1, 1], [1, 2], [2, 2], [3, 0], [3, 2]], "position": [1, 3], "goal": [2, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.68, "north": 0.32}, "solvable": {"false": 0.63, "true": 0.37}, "value": {"0": 0.28, "1": 0.21, "2": 0.11, "3": 0.4}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:66fbf43dc6bff2674ced6463", "state_id": "tic_tac_toe:66fbf43dc6bff2674ced6463", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X .\nO . .\nX . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place X in row 1, column 3 (cell 3).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_3": 0.0, "cell_9": 0.25, "cell_5": 0.25, "cell_8": 0.25, "cell_6": 0.25}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5", "cell_6", "cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.....XXOO:X", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OX.O..X..", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.56, "cell_3": 0.15, "cell_6": 0.09, "cell_8": 0.11, "cell_9": 0.09}, "solvable": {"false": 0.78, "true": 0.22}, "value": {"0": 0.21, "1": 0.52, "2": 0.27}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:47c390db0a20c5dcf8538f3a", "state_id": "grid_navigation:47c390db0a20c5dcf8538f3a", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # .\n# . # .\nA . # .\n. . G #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.5, "south": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (1, 1), (2, 1), (2, 3), (3, 1), (3, 3))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [1, 0], [1, 2], [2, 2], [3, 3]], "position": [2, 0], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.67, "east": 0.33}, "solvable": {"false": 0.15000000000000002, "true": 0.85}, "value": {"0": 0.2, "1": 0.65, "2": 0.09, "3": 0.06}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:7dd83a5a00cebfe9f187bd67", "state_id": "grid_navigation:7dd83a5a00cebfe9f187bd67", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG . . .\n. . # #\nA . . #\n. . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.0, "north": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((1, 2), (1, 3), (2, 3), (3, 2))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 2], [1, 3], [2, 3], [3, 2]], "position": [2, 0], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.52, "east": 0.43, "south": 0.05}, "solvable": {"false": 0.22999999999999998, "true": 0.77}, "value": {"0": 0.08, "1": 0.68, "2": 0.16, "3": 0.08}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:e023dd571222709b531218f2", "state_id": "tic_tac_toe:e023dd571222709b531218f2", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X X\n. . .\nX . O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place O in row 3, column 2 (cell 8).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_6": "Place O in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_8": 0.0, "cell_4": 0.0, "cell_5": 1.0, "cell_6": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.X...XXO:O", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXX...X.O", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.22, "cell_4": 0.21, "cell_6": 0.09, "cell_5": 0.48}, "solvable": {"false": 0.89, "true": 0.11}, "value": {"0": 0.56, "1": 0.3, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:78e7699f827105811fa2998f", "state_id": "grid_navigation:78e7699f827105811fa2998f", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # #\nG # . #\n. . . .\n. # A .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (1, 1), (1, 3), (2, 0), (3, 0), (3, 1))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [0, 3], [1, 1], [1, 3], [3, 1]], "position": [3, 2], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.69, "east": 0.31}, "solvable": {"false": 0.41000000000000003, "true": 0.59}, "value": {"0": 0.01, "1": 0.44, "2": 0.41, "3": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:97f479ae2c5dad694499d082", "state_id": "grid_navigation:97f479ae2c5dad694499d082", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # . A\n# . . .\n. . . G\n# . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "south": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (2, 3), (3, 0), (3, 2), (3, 3))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [1, 0], [3, 0], [3, 3]], "position": [0, 3], "goal": [2, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.85, "west": 0.15}, "solvable": {"false": 0.07999999999999996, "true": 0.92}, "value": {"0": 0.04, "1": 0.82, "2": 0.12, "3": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:f88f9d05556450201e16127d", "state_id": "grid_navigation:f88f9d05556450201e16127d", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG . # .\n. # . #\n# . A .\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north.", "west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"south": 0.25, "north": 0.25, "west": 0.25, "east": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north", "south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (1, 1), (1, 3), (2, 0))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 1], [1, 3], [2, 0]], "position": [2, 2], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.21, "north": 0.27, "south": 0.24, "west": 0.28}, "solvable": {"false": 0.44999999999999996, "true": 0.55}, "value": {"0": 0.03, "1": 0.42, "2": 0.37, "3": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:393d904562a27c2526d55962", "state_id": "tic_tac_toe:393d904562a27c2526d55962", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . O\nX . X\n. X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.0, "cell_2": 0.5, "cell_5": 0.5, "cell_9": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.X.X.XO.O:O", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.OX.X.X.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.15, "cell_2": 0.24, "cell_5": 0.48, "cell_9": 0.13}, "solvable": {"false": 0.83, "true": 0.17}, "value": {"0": 0.39, "1": 0.37, "2": 0.24}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:030400c5c64403612b0fd25c", "state_id": "grid_navigation:030400c5c64403612b0fd25c", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n. # # .\nG . . A\n. . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.0, "west": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 2), (2, 0), (2, 2), (2, 3))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 1], [1, 2], [3, 2]], "position": [2, 3], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.05, "west": 0.85, "north": 0.1}, "solvable": {"false": 0.25, "true": 0.75}, "value": {"0": 0.1, "1": 0.56, "2": 0.25, "3": 0.09}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:0887e47be60afe5fc28440f4", "state_id": "grid_navigation:0887e47be60afe5fc28440f4", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. . . .\n. . . G\n# . A .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "east": 0.5, "north": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((3, 0),)", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[3, 0]], "position": [3, 2], "goal": [2, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.11, "west": 0.02, "north": 0.87}, "solvable": {"false": 0.06999999999999995, "true": 0.93}, "value": {"0": 0.63, "1": 0.36, "2": 0.01, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:981a568c91319b3b25313aa3", "state_id": "tic_tac_toe:981a568c91319b3b25313aa3", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . O\nO X X\nX O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place O in row 1, column 2 (cell 2).", "cell_1": "Place O in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_2": 0.0, "cell_1": 1.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OOXXXOX:O", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..OOXXXOX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.66, "cell_2": 0.34}, "solvable": {"false": 0.88, "true": 0.12}, "value": {"0": 0.53, "1": 0.33, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:7850aacdcb5d0b6c6d884a6b", "state_id": "tic_tac_toe:7850aacdcb5d0b6c6d884a6b", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . X\nX O .\nO X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_1": 1.0, "cell_2": 0.0, "cell_6": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XXO.OXO:X", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..XXO.OXO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.46, "cell_2": 0.25, "cell_1": 0.29}, "solvable": {"false": 0.74, "true": 0.26}, "value": {"0": 0.23, "1": 0.49, "2": 0.28}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:f863b5c7769f32c0e604d862", "state_id": "tic_tac_toe:f863b5c7769f32c0e604d862", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O X\n. . .\n. O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place O in row 2, column 3 (cell 6).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.0, "cell_4": 0.0, "cell_7": 0.0, "cell_5": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XO.OX.X:O", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOX....OX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.16, "cell_7": 0.16, "cell_4": 0.17, "cell_5": 0.51}, "solvable": {"false": 0.9, "true": 0.1}, "value": {"0": 0.49, "1": 0.38, "2": 0.13}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:11fc0040b68aefd60229cda6", "state_id": "grid_navigation:11fc0040b68aefd60229cda6", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nA . # #\n. # . .\n# . # G\n. # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"south": 0.5, "east": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 1), (1, 3), (2, 0), (2, 2), (3, 1))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [1, 1], [2, 0], [2, 2], [3, 1]], "position": [0, 0], "goal": [2, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.21, "south": 0.79}, "solvable": {"false": 0.42000000000000004, "true": 0.58}, "value": {"0": 0.01, "1": 0.32, "2": 0.47, "3": 0.2}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:9fd825a0229ffa54e32655ec", "state_id": "grid_navigation:9fd825a0229ffa54e32655ec", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . .\n. . . #\nA . # .\n. # . G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.3333333333333333, "south": 0.3333333333333333, "north": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (1, 1), (2, 0), (3, 3))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 3], [2, 2], [3, 1]], "position": [2, 0], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.59, "south": 0.3, "north": 0.11}, "solvable": {"false": 0.29000000000000004, "true": 0.71}, "value": {"0": 0.01, "1": 0.58, "2": 0.29, "3": 0.12}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:6f8fa587658259e058ee12fa", "state_id": "tic_tac_toe:6f8fa587658259e058ee12fa", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O O\n. . O\nX . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place X in row 3, column 2 (cell 8).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_8": 0.3333333333333333, "cell_4": 0.3333333333333333, "cell_5": 0.3333333333333333}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4", "cell_5", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:OOXO..X.X:X", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOO..OX.X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.32, "cell_5": 0.45, "cell_4": 0.23}, "solvable": {"false": 0.6, "true": 0.4}, "value": {"0": 0.31, "1": 0.31, "2": 0.38}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:d9285991fa057fa0fd23cf5e", "state_id": "grid_navigation:d9285991fa057fa0fd23cf5e", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # # .\n. # . #\n. # G #\n. . A .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "north": 1.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (0, 2), (1, 3), (2, 1), (2, 2), (2, 3), (3, 3))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [0, 2], [1, 1], [1, 3], [2, 1], [2, 3]], "position": [3, 2], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.69, "east": 0.11, "west": 0.2}, "solvable": {"false": 0.4, "true": 0.6}, "value": {"0": 0.21, "1": 0.46, "2": 0.15, "3": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:6c8c86365a0c5b7aebddd069", "state_id": "grid_navigation:6c8c86365a0c5b7aebddd069", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . .\n. # . #\n# . # .\nG . . A", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 1.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 3), (1, 2), (2, 1), (3, 2))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 1], [1, 3], [2, 0], [2, 2]], "position": [3, 3], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.8, "north": 0.2}, "solvable": {"false": 0.28, "true": 0.72}, "value": {"0": 0.04, "1": 0.51, "2": 0.3, "3": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:842e89c6a1acd76b9741693f", "state_id": "grid_navigation:842e89c6a1acd76b9741693f", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . #\nA . . #\n. . # .\nG . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 1.0, "north": 0.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 3), (1, 2), (2, 3), (3, 3))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 3], [2, 2], [3, 3]], "position": [1, 0], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.05, "south": 0.94, "north": 0.01}, "solvable": {"false": 0.22999999999999998, "true": 0.77}, "value": {"0": 0.31, "1": 0.52, "2": 0.09, "3": 0.08}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:90ffe2e2caa087ce162d235c", "state_id": "tic_tac_toe:90ffe2e2caa087ce162d235c", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O .\nO . .\n. X X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place X in row 2, column 2 (cell 5).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 0.2, "cell_3": 0.2, "cell_1": 0.2, "cell_7": 0.2, "cell_6": 0.2}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_3", "cell_5", "cell_6", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XO.X.O.:X", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".O.O...XX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.09, "cell_7": 0.2, "cell_5": 0.38, "cell_1": 0.13, "cell_6": 0.2}, "solvable": {"false": 0.6599999999999999, "true": 0.34}, "value": {"0": 0.23, "1": 0.4, "2": 0.37}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:35a1a2b69465f9b0f491b047", "state_id": "tic_tac_toe:35a1a2b69465f9b0f491b047", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . X\n. . X\nO . O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place O in row 2, column 2 (cell 5).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_8", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 0.0, "cell_8": 1.0, "cell_4": 0.0, "cell_2": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.O..XX.X:O", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.X..XO.O", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.17, "cell_8": 0.18, "cell_5": 0.5, "cell_4": 0.15}, "solvable": {"false": 0.83, "true": 0.17}, "value": {"0": 0.41, "1": 0.34, "2": 0.25}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:31c8ba417b75819827385307", "state_id": "tic_tac_toe:31c8ba417b75819827385307", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO O X\n. X X\n. . O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place X in row 3, column 2 (cell 8).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_4": "Place X in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_8": 0.3333333333333333, "cell_7": 0.3333333333333333, "cell_4": 0.3333333333333333}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4", "cell_7", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O.XOOXX:X", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OOX.XX..O", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.39, "cell_7": 0.34, "cell_8": 0.27}, "solvable": {"false": 0.6, "true": 0.4}, "value": {"0": 0.17, "1": 0.39, "2": 0.44}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:f8a981eec8f98de2dbfe9eb2", "state_id": "grid_navigation:f8a981eec8f98de2dbfe9eb2", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\nG A # #\n# . . #\n# # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "north": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (0, 3), (1, 3), (2, 1), (2, 3), (3, 1), (3, 2))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 2], [1, 3], [2, 0], [2, 3], [3, 0], [3, 1], [3, 2]], "position": [1, 1], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.95, "north": 0.03, "south": 0.02}, "solvable": {"false": 0.19999999999999996, "true": 0.8}, "value": {"0": 0.93, "1": 0.03, "2": 0.01, "3": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:7c9003589da1b215c03ff947", "state_id": "tic_tac_toe:7c9003589da1b215c03ff947", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . O\nO O X\n. . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place X in row 1, column 2 (cell 2).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_7", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_2": 0.0, "cell_8": 0.0, "cell_7": 1.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOOXX.O:X", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.OOOX..X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.25, "cell_7": 0.42, "cell_2": 0.33}, "solvable": {"false": 0.63, "true": 0.37}, "value": {"0": 0.36, "1": 0.35, "2": 0.29}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:243741e9a9e50e80fc283bef", "state_id": "tic_tac_toe:243741e9a9e50e80fc283bef", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . .\nO X O\n. . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_3": "Place X in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.2, "cell_2": 0.2, "cell_7": 0.2, "cell_8": 0.2, "cell_3": 0.2}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_3", "cell_7", "cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...OXO..X:X", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X..OXO...", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.32, "cell_3": 0.2, "cell_7": 0.18, "cell_9": 0.11, "cell_8": 0.19}, "solvable": {"false": 0.8200000000000001, "true": 0.18}, "value": {"0": 0.16, "1": 0.6, "2": 0.24}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:9e731bed222dff349d38699b", "state_id": "tic_tac_toe:9e731bed222dff349d38699b", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X O\n. X X\nO . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place O in row 3, column 2 (cell 8).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_8": 0.25, "cell_4": 0.25, "cell_1": 0.25, "cell_9": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_4", "cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXX.OX.:O", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XO.XXO..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.4, "cell_8": 0.32, "cell_9": 0.09, "cell_1": 0.19}, "solvable": {"false": 0.84, "true": 0.16}, "value": {"0": 0.55, "1": 0.24, "2": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:0e4b038447550f7ab71e9327", "state_id": "grid_navigation:0e4b038447550f7ab71e9327", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # .\n. # . .\nG # # #\nA . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 1.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 1), (1, 2), (2, 1), (2, 3), (3, 1))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [1, 1], [2, 1], [2, 2], [2, 3]], "position": [3, 0], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.53, "east": 0.47}, "solvable": {"false": 0.5800000000000001, "true": 0.42}, "value": {"0": 0.04, "1": 0.27, "2": 0.29, "3": 0.4}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:36b07212a128ffe33a0859cb", "state_id": "tic_tac_toe:36b07212a128ffe33a0859cb", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . O\n. . O\nX O X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place X in row 2, column 2 (cell 5).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 0.3333333333333333, "cell_4": 0.3333333333333333, "cell_2": 0.3333333333333333}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_4", "cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.XO..XOX:X", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.O..OXOX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.28, "cell_5": 0.4, "cell_2": 0.32}, "solvable": {"false": 0.75, "true": 0.25}, "value": {"0": 0.25, "1": 0.49, "2": 0.26}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:e6fcd09940f363badb1bff5c", "state_id": "tic_tac_toe:e6fcd09940f363badb1bff5c", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X X\nO . .\nX X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place O in row 2, column 3 (cell 6).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.0, "cell_5": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.XX.XXOO:O", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXXO..XXO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.77, "cell_6": 0.23}, "solvable": {"false": 0.8200000000000001, "true": 0.18}, "value": {"0": 0.53, "1": 0.26, "2": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:8d726b00c2e0eccb8d3e0a6a", "state_id": "grid_navigation:8d726b00c2e0eccb8d3e0a6a", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # .\n# . # .\n. . G #\n. # A .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 1.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (1, 2), (1, 3), (2, 0), (2, 3), (3, 2))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [1, 0], [1, 2], [2, 3], [3, 1]], "position": [3, 2], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.69, "east": 0.31}, "solvable": {"false": 0.26, "true": 0.74}, "value": {"0": 0.33, "1": 0.47, "2": 0.09, "3": 0.11}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:98f63e1e41d3063e72278775", "state_id": "grid_navigation:98f63e1e41d3063e72278775", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\n. G . #\n. A . .\n. . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east.", "west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 1.0, "east": 0.0, "west": 0.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (1, 3), (3, 2), (3, 3))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 3], [3, 2], [3, 3]], "position": [2, 1], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.88, "east": 0.05, "west": 0.06, "south": 0.01}, "solvable": {"false": 0.13, "true": 0.87}, "value": {"0": 0.8, "1": 0.15, "2": 0.03, "3": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:6adf6a64eb9edf45c3022582", "state_id": "grid_navigation:6adf6a64eb9edf45c3022582", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . G .\n# . . #\n. . . .\n. # A .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 1.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (1, 3), (3, 2))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [1, 3], [3, 1]], "position": [3, 2], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.24, "north": 0.76}, "solvable": {"false": 0.14, "true": 0.86}, "value": {"0": 0.01, "1": 0.58, "2": 0.35, "3": 0.06}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:f8e12d2bcb72fa902dfb0e18", "state_id": "tic_tac_toe:f8e12d2bcb72fa902dfb0e18", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O .\n. O X\nX . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_8": "Place X in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_8", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_4": 0.0, "cell_3": 0.0, "cell_1": 0.0, "cell_9": 0.0, "cell_8": 1.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOO..X.:X", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".O..OXX..", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.28, "cell_3": 0.11, "cell_1": 0.18, "cell_8": 0.31, "cell_9": 0.12}, "solvable": {"false": 0.76, "true": 0.24}, "value": {"0": 0.22, "1": 0.45, "2": 0.33}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:a874c183df9ec93fd5034c12", "state_id": "grid_navigation:a874c183df9ec93fd5034c12", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n. . . A\n. . . #\nG . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 1.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 3), (1, 3), (3, 2))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [2, 3], [3, 3]], "position": [1, 3], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.04, "west": 0.96}, "solvable": {"false": 0.17000000000000004, "true": 0.83}, "value": {"0": 0.02, "1": 0.64, "2": 0.26, "3": 0.08}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:ab98273ef7fbf63db8f9689b", "state_id": "grid_navigation:ab98273ef7fbf63db8f9689b", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# A . #\n# . # .\n. . # .\n# # G #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"south": 0.5, "east": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 2), (0, 3), (1, 1), (2, 1), (2, 3), (3, 0), (3, 3))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 3], [1, 0], [1, 2], [2, 2], [3, 0], [3, 1], [3, 3]], "position": [0, 1], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.05, "south": 0.95}, "solvable": {"false": 0.47, "true": 0.53}, "value": {"0": 0.02, "1": 0.49, "2": 0.27, "3": 0.22}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:81151df4114712809014ceaf", "state_id": "grid_navigation:81151df4114712809014ceaf", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG # . .\n. . . .\n# A . #\n. . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.0, "south": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (2, 0), (2, 3), (3, 2), (3, 3))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [2, 0], [2, 3], [3, 2], [3, 3]], "position": [2, 1], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.73, "east": 0.21, "south": 0.06}, "solvable": {"false": 0.30000000000000004, "true": 0.7}, "value": {"0": 0.07, "1": 0.58, "2": 0.22, "3": 0.13}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:684dbadb9388895e828fa12f", "state_id": "tic_tac_toe:684dbadb9388895e828fa12f", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . X\n. O X\n. X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.0, "cell_9": 1.0, "cell_4": 0.0, "cell_2": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXO..XX:O", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.X.OX.X.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.41, "cell_2": 0.09, "cell_7": 0.3, "cell_4": 0.2}, "solvable": {"false": 0.8200000000000001, "true": 0.18}, "value": {"0": 0.5, "1": 0.29, "2": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:cb8b24a2e1e24dea9740187f", "state_id": "tic_tac_toe:cb8b24a2e1e24dea9740187f", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . O\nO O X\n. X X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_7", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 1.0, "cell_2": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXXO.XXO:O", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.OOOX.XX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.36, "cell_7": 0.64}, "solvable": {"false": 0.75, "true": 0.25}, "value": {"0": 0.31, "1": 0.3, "2": 0.39}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:a1bdcde50d022c4b4f107d79", "state_id": "tic_tac_toe:a1bdcde50d022c4b4f107d79", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X O\nO . O\nX X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.0, "cell_5": 0.5, "cell_9": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOX.XXO.:X", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XOO.OXX.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.16, "cell_5": 0.42, "cell_9": 0.42}, "solvable": {"false": 0.53, "true": 0.47}, "value": {"0": 0.35, "1": 0.24, "2": 0.41}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:64eb30b68428cf66a6921504", "state_id": "grid_navigation:64eb30b68428cf66a6921504", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. G . .\n. . . #\n. . . .\nA . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.5, "north": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 3),)", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 3]], "position": [3, 0], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.1, "north": 0.9}, "solvable": {"false": 0.07999999999999996, "true": 0.92}, "value": {"0": 0.02, "1": 0.7, "2": 0.26, "3": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:f8fc6af6c8405c9ed23d1619", "state_id": "tic_tac_toe:f8fc6af6c8405c9ed23d1619", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X .\n. X O\nO . X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place O in row 1, column 3 (cell 3).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_1": "Place O in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_3": 0.25, "cell_4": 0.25, "cell_8": 0.25, "cell_1": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_3", "cell_4", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXX..OX:O", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".X..XOO.X", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.24, "cell_3": 0.16, "cell_8": 0.42, "cell_1": 0.18}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.42, "1": 0.39, "2": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:a14a434856229e0c82b3b137", "state_id": "tic_tac_toe:a14a434856229e0c82b3b137", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . X\n. . .\nX . O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place X in row 1, column 2 (cell 2).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.0, "cell_8": 0.0, "cell_4": 0.0, "cell_5": 1.0, "cell_6": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.X...X.O:X", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.X...X.O", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.08, "cell_4": 0.14, "cell_5": 0.55, "cell_2": 0.11, "cell_8": 0.12}, "solvable": {"false": 0.75, "true": 0.25}, "value": {"0": 0.13, "1": 0.48, "2": 0.39}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:fa0348f44a86413dfaa91c6c", "state_id": "tic_tac_toe:fa0348f44a86413dfaa91c6c", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X .\nX O X\nO O .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.5, "cell_3": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOXOX.XX:O", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XX.XOXOO.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.44, "cell_3": 0.56}, "solvable": {"false": 0.78, "true": 0.22}, "value": {"0": 0.49, "1": 0.25, "2": 0.26}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:9fa572fe61167e4745dc2eaf", "state_id": "grid_navigation:9fa572fe61167e4745dc2eaf", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # G\n# # # .\n# . # #\n. . A .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "west": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (1, 1), (1, 2), (1, 3), (2, 0), (2, 1), (2, 3))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 0], [1, 1], [1, 2], [2, 0], [2, 2], [2, 3]], "position": [3, 2], "goal": [0, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.48, "west": 0.52}, "solvable": {"false": 0.52, "true": 0.48}, "value": {"0": 0.02, "1": 0.28, "2": 0.37, "3": 0.33}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:9656b752a24745730ca2e040", "state_id": "tic_tac_toe:9656b752a24745730ca2e040", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X .\nX O O\n. . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place X in row 3, column 2 (cell 8).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_8": 0.0, "cell_3": 0.0, "cell_1": 1.0, "cell_9": 0.0, "cell_7": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...OOX.X.:X", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".X.XOO...", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.11, "cell_3": 0.18, "cell_1": 0.36, "cell_8": 0.15, "cell_7": 0.2}, "solvable": {"false": 0.77, "true": 0.23}, "value": {"0": 0.18, "1": 0.48, "2": 0.34}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:0d02a7926a6fbb85d4396810", "state_id": "grid_navigation:0d02a7926a6fbb85d4396810", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\n. A . #\n# . . .\n. . # G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.5, "east": 0.5, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (1, 3), (2, 0), (3, 2))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 3], [2, 0], [3, 2]], "position": [1, 1], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.29, "west": 0.04, "south": 0.67}, "solvable": {"false": 0.21999999999999997, "true": 0.78}, "value": {"0": 0.01, "1": 0.63, "2": 0.26, "3": 0.1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:4157210fe94f77c0810d7f96", "state_id": "tic_tac_toe:4157210fe94f77c0810d7f96", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X O\nX . .\n. O .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place O in row 2, column 3 (cell 6).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_7": "Place O in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_7", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.0, "cell_9": 0.0, "cell_5": 0.0, "cell_7": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OO.X.XX:O", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XXOX...O.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.57, "cell_7": 0.15, "cell_6": 0.16, "cell_9": 0.12}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.65, "1": 0.2, "2": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:374c826550604e2055b6eb80", "state_id": "grid_navigation:374c826550604e2055b6eb80", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # #\n# . . G\n. # . .\nA . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (1, 0), (1, 3), (2, 2), (3, 1))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [1, 0], [2, 1], [3, 2]], "position": [3, 0], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.46, "east": 0.54}, "solvable": {"false": 0.27, "true": 0.73}, "value": {"0": 0.01, "1": 0.51, "2": 0.35, "3": 0.13}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:17e8e726a120a22ae5fb76be", "state_id": "tic_tac_toe:17e8e726a120a22ae5fb76be", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O X\n. X X\nO . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_1": 0.25, "cell_4": 0.25, "cell_8": 0.25, "cell_9": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_4", "cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OOX.XX.:O", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".OX.XXO..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.19, "cell_8": 0.23, "cell_4": 0.45, "cell_9": 0.13}, "solvable": {"false": 0.88, "true": 0.12}, "value": {"0": 0.71, "1": 0.15, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:59c9aca70e5afa5360b21336", "state_id": "grid_navigation:59c9aca70e5afa5360b21336", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # #\n. . . G\n. . . .\n# . . A", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 1.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (1, 0), (3, 0), (3, 3))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [0, 3], [3, 0]], "position": [3, 3], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.24, "north": 0.76}, "solvable": {"false": 0.09999999999999998, "true": 0.9}, "value": {"0": 0.02, "1": 0.67, "2": 0.28, "3": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:8a54110c7cc3948717955b11", "state_id": "tic_tac_toe:8a54110c7cc3948717955b11", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO O .\n. . X\nX X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_4": 0.3333333333333333, "cell_3": 0.3333333333333333, "cell_5": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_4", "cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOX..OXX:X", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OO...XXXO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.27, "cell_3": 0.41, "cell_5": 0.32}, "solvable": {"false": 0.51, "true": 0.49}, "value": {"0": 0.15, "1": 0.29, "2": 0.56}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:11a75dc59ca54e5032ab943d", "state_id": "tic_tac_toe:11a75dc59ca54e5032ab943d", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O .\nX X .\n. O O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_3": "Place X in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_6", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.5, "cell_6": 0.5, "cell_3": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_6", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OOXOXX.:X", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XO.XX..OO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.44, "cell_3": 0.26, "cell_7": 0.3}, "solvable": {"false": 0.54, "true": 0.46}, "value": {"0": 0.23, "1": 0.39, "2": 0.38}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:8826fb34d0db6b2d9d90f5df", "state_id": "tic_tac_toe:8826fb34d0db6b2d9d90f5df", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O O\nX . O\n. . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place X in row 3, column 2 (cell 8).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_8": 0.3333333333333333, "cell_7": 0.3333333333333333, "cell_5": 0.3333333333333333}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5", "cell_7", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XX.OXOO:X", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOOX.O..X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.26, "cell_5": 0.54, "cell_8": 0.2}, "solvable": {"false": 0.6799999999999999, "true": 0.32}, "value": {"0": 0.2, "1": 0.49, "2": 0.31}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:1cd306a81517c5fdbbad662e", "state_id": "grid_navigation:1cd306a81517c5fdbbad662e", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG . . .\nA . . .\n. . . .\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.0, "north": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=()", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [], "position": [1, 0], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.92, "east": 0.08, "south": 0}, "solvable": {"false": 0.10999999999999999, "true": 0.89}, "value": {"0": 0.78, "1": 0.2, "2": 0.02, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:415e56b135fe4a678d5c4bc2", "state_id": "grid_navigation:415e56b135fe4a678d5c4bc2", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. A . .\n# . G .\n. # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north.", "south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.5, "north": 0.0, "south": 0.5, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (0, 2), (1, 3))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[2, 0], [3, 1], [3, 2]], "position": [1, 1], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.02, "south": 0.22, "north": 0.02, "east": 0.74}, "solvable": {"false": 0.13, "true": 0.87}, "value": {"0": 0.76, "1": 0.18, "2": 0.03, "3": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:dd9e2d79676e8de73ed8d5b3", "state_id": "grid_navigation:dd9e2d79676e8de73ed8d5b3", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG # . .\n. . . #\nA . . #\n. # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.0, "south": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (1, 3), (2, 3), (3, 1), (3, 2))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 3], [2, 3], [3, 1], [3, 2]], "position": [2, 0], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.51, "south": 0.04, "east": 0.45}, "solvable": {"false": 0.33999999999999997, "true": 0.66}, "value": {"0": 0.04, "1": 0.56, "2": 0.22, "3": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:293f0b97cb0674e51cca64ef", "state_id": "tic_tac_toe:293f0b97cb0674e51cca64ef", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . X\nO O X\nX X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 1.0, "cell_2": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.XXXO.XOO:O", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.XOOXXX.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.15, "cell_9": 0.85}, "solvable": {"false": 0.74, "true": 0.26}, "value": {"0": 0.44, "1": 0.31, "2": 0.25}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:63ae801fab21982a936fa466", "state_id": "tic_tac_toe:63ae801fab21982a936fa466", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O X\nO X X\nO . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_8": "Place X in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.0, "cell_9": 1.0, "cell_8": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXXOXO.:X", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".OXOXXO..", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.23, "cell_8": 0.57, "cell_9": 0.2}, "solvable": {"false": 0.55, "true": 0.45}, "value": {"0": 0.41000000000000003, "1": 0.23, "2": 0.36}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:dd43b0e43a0e43b6162359d3", "state_id": "grid_navigation:dd43b0e43a0e43b6162359d3", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # G\n# . # #\n. A . .\n# . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south.", "east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.25, "south": 0.25, "east": 0.25, "north": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north", "south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 2), (1, 0), (1, 1), (1, 3), (3, 0), (3, 3))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [1, 0], [1, 2], [1, 3], [3, 0], [3, 3]], "position": [2, 1], "goal": [0, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.17, "south": 0.05, "north": 0.56, "east": 0.21}, "solvable": {"false": 0.62, "true": 0.38}, "value": {"0": 0.04, "1": 0.34, "2": 0.21, "3": 0.41}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:6166f55689f3763259c162e4", "state_id": "tic_tac_toe:6166f55689f3763259c162e4", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . X\nO O .\n. O X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place X in row 2, column 3 (cell 6).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.5, "cell_2": 0.5, "cell_7": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXOO.X.X:X", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.XOO..OX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.43, "cell_7": 0.27, "cell_2": 0.3}, "solvable": {"false": 0.55, "true": 0.45}, "value": {"0": 0.16, "1": 0.37, "2": 0.47000000000000003}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:658a5dcbf439200f95c6c2cd", "state_id": "grid_navigation:658a5dcbf439200f95c6c2cd", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . A .\n. . . .\n# . G #\n# . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "east": 0.0, "south": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (3, 0), (3, 1))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[2, 0], [2, 3], [3, 0]], "position": [0, 2], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.19, "west": 0.07, "south": 0.74}, "solvable": {"false": 0.07999999999999996, "true": 0.92}, "value": {"0": 0.03, "1": 0.79, "2": 0.15, "3": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:b041a6bb722831da890b3780", "state_id": "grid_navigation:b041a6bb722831da890b3780", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n. . G .\n. . . #\n# # A .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (2, 0), (3, 2), (3, 3))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [2, 3], [3, 0], [3, 1]], "position": [3, 2], "goal": [1, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.13, "north": 0.87}, "solvable": {"false": 0.14, "true": 0.86}, "value": {"0": 0.05, "1": 0.74, "2": 0.17, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:9a1b70a9f4a70efda73597bb", "state_id": "tic_tac_toe:9a1b70a9f4a70efda73597bb", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . .\n. O X\nX . O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place O in row 2, column 1 (cell 4).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_4": 1.0, "cell_3": 0.0, "cell_8": 0.0, "cell_2": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XXO.O.X:O", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X...OXX.O", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.11, "cell_8": 0.38, "cell_2": 0.2, "cell_4": 0.31}, "solvable": {"false": 0.8200000000000001, "true": 0.18}, "value": {"0": 0.39, "1": 0.38, "2": 0.23}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:1d4b5acd5c5484393c800192", "state_id": "grid_navigation:1d4b5acd5c5484393c800192", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # #\nA # . .\nG # # #\n. . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "south": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 1), (1, 2), (2, 0), (2, 1), (2, 3), (3, 1), (3, 3))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [1, 1], [2, 1], [2, 2], [2, 3], [3, 2]], "position": [1, 0], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.23, "south": 0.77}, "solvable": {"false": 0.73, "true": 0.27}, "value": {"0": 0.33, "1": 0.18, "2": 0.11, "3": 0.38}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:c1bbaf350e1649c27e9b33f0", "state_id": "grid_navigation:c1bbaf350e1649c27e9b33f0", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # G #\n. # # #\n. . A .\n# # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "west": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 2), (1, 0), (1, 1), (1, 2), (3, 1), (3, 2), (3, 3))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 3], [1, 1], [1, 2], [1, 3], [3, 0], [3, 1], [3, 2]], "position": [2, 2], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.37, "west": 0.63}, "solvable": {"false": 0.74, "true": 0.26}, "value": {"0": 0.04, "1": 0.17, "2": 0.14, "3": 0.65}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:9df7eebf35dcfc0a0c49411d", "state_id": "tic_tac_toe:9df7eebf35dcfc0a0c49411d", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X X\n. O .\nO X X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place O in row 2, column 3 (cell 6).", "cell_4": "Place O in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.0, "cell_4": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.OXOXX.X:O", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXX.O.OXX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.3, "cell_4": 0.7}, "solvable": {"false": 0.86, "true": 0.14}, "value": {"0": 0.51, "1": 0.29, "2": 0.2}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:1df6e047ebf15c920b0a7280", "state_id": "grid_navigation:1df6e047ebf15c920b0a7280", "family_id": "grid_navigation_bfs_v1", "split": "calibration", "state": "Recorded game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG . # .\n. # # #\nA . . #\n. . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "north": 1.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (1, 1), (1, 2), (1, 3), (2, 3), (3, 3))", "template_id": "grid_navigation:calibration:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 1], [1, 2], [1, 3], [2, 3], [3, 3]], "position": [2, 0], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.28, "south": 0.13, "north": 0.59}, "solvable": {"false": 0.51, "true": 0.49}, "value": {"0": 0.03, "1": 0.37, "2": 0.34, "3": 0.26}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:78a23a889db4a51df4e010f1", "state_id": "tic_tac_toe:78a23a889db4a51df4e010f1", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\nX . O\n. X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place X in row 1, column 2 (cell 2).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.0, "cell_1": 0.0, "cell_3": 1.0, "cell_5": 0.0, "cell_7": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...O.XOX.:X", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "...X.O.XO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.19, "cell_1": 0.2, "cell_2": 0.19, "cell_5": 0.1, "cell_7": 0.32}, "solvable": {"false": 0.6699999999999999, "true": 0.33}, "value": {"0": 0.37, "1": 0.35, "2": 0.28}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:c541e7c7cf2a9666a9e0ed93", "state_id": "tic_tac_toe:c541e7c7cf2a9666a9e0ed93", "family_id": "tic_tac_toe_minimax_v1", "split": "calibration", "state": "Recorded game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X X\nO X O\n. . X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_8": "Place O in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_7", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 1.0, "cell_8": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOXOOXX:O", "template_id": "tic_tac_toe:calibration:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXXOXO..X", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.35, "cell_7": 0.65}, "solvable": {"false": 0.88, "true": 0.12}, "value": {"0": 0.5, "1": 0.38, "2": 0.12}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:eb1fd0cf3ac37b13cda0fd47", "state_id": "grid_navigation:eb1fd0cf3ac37b13cda0fd47", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\n. A . #\n# . . #\n# G . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "south": 1.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (1, 0), (1, 3), (2, 3), (3, 1))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 3], [2, 0], [2, 3], [3, 0]], "position": [1, 1], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.8099999999999999, "west": 0.16, "east": 0.03}, "solvable": {"false": 0.17000000000000004, "true": 0.83}, "value": {"0": 0.1, "1": 0.68, "2": 0.13, "3": 0.09}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:2c7f9bd3b8edfab42a2a3fcd", "state_id": "grid_navigation:2c7f9bd3b8edfab42a2a3fcd", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\n. # # #\n# # . .\nG . A #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 1.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (1, 1), (1, 2), (1, 3), (2, 2), (3, 0), (3, 2))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 1], [1, 2], [1, 3], [2, 0], [2, 1], [3, 3]], "position": [3, 2], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.88, "north": 0.12}, "solvable": {"false": 0.39, "true": 0.61}, "value": {"0": 0.29, "1": 0.41000000000000003, "2": 0.22, "3": 0.08}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:bdedc2f28f54f0688eaecd84", "state_id": "tic_tac_toe:bdedc2f28f54f0688eaecd84", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X .\nO . .\nX O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place O in row 1, column 3 (cell 3).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_6": "Place O in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_3": 0.5, "cell_1": 0.0, "cell_5": 0.5, "cell_6": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XX.O.OX:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".X.O..XOX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.19, "cell_6": 0.14, "cell_3": 0.14, "cell_5": 0.53}, "solvable": {"false": 0.89, "true": 0.11}, "value": {"0": 0.51, "1": 0.31, "2": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:c9bbc70203514e27fd83e312", "state_id": "tic_tac_toe:c9bbc70203514e27fd83e312", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . .\n. . O\n. O X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place X in row 1, column 3 (cell 3).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_3": 0.2, "cell_5": 0.2, "cell_7": 0.2, "cell_4": 0.2, "cell_2": 0.2}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_3", "cell_4", "cell_5", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XO..XO.:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X....O.OX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.21, "cell_4": 0.16, "cell_5": 0.39, "cell_3": 0.11, "cell_2": 0.13}, "solvable": {"false": 0.76, "true": 0.24}, "value": {"0": 0.29, "1": 0.42, "2": 0.29}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:7fa0b54d2ac88605083825b5", "state_id": "tic_tac_toe:7fa0b54d2ac88605083825b5", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X X\n. . O\nX . O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place X in row 2, column 2 (cell 5).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_4": "Place X in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 1.0, "cell_8": 0.0, "cell_4": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.XO..XXO:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXX..OX.O", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.35, "cell_4": 0.23, "cell_5": 0.42}, "solvable": {"false": 0.55, "true": 0.45}, "value": {"0": 0.23, "1": 0.36, "2": 0.41}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:658d381a3a0de76d2b2c600e", "state_id": "grid_navigation:658d381a3a0de76d2b2c600e", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # #\n. . # #\n. # . .\nA . G .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 2), (2, 0), (2, 1), (3, 0), (3, 1))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [1, 2], [1, 3], [2, 1]], "position": [3, 0], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.25, "east": 0.75}, "solvable": {"false": 0.26, "true": 0.74}, "value": {"0": 0.35, "1": 0.43, "2": 0.15, "3": 0.07}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:254ae4423e957538ebd82e01", "state_id": "tic_tac_toe:254ae4423e957538ebd82e01", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\nX X O\n. . O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place X in row 1, column 3 (cell 3).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_3": 1.0, "cell_7": 0.0, "cell_1": 0.0, "cell_8": 0.0, "cell_2": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...OXXO..:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "...XXO..O", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.1, "cell_8": 0.28, "cell_7": 0.21, "cell_3": 0.24, "cell_1": 0.17}, "solvable": {"false": 0.77, "true": 0.23}, "value": {"0": 0.39999999999999997, "1": 0.39, "2": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:8809599093a4f0e14d72ab71", "state_id": "tic_tac_toe:8809599093a4f0e14d72ab71", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . X\nO O X\n. X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_1": 1.0, "cell_2": 0.0, "cell_7": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOOX.XO:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..XOOX.XO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.26, "cell_7": 0.45, "cell_1": 0.29}, "solvable": {"false": 0.52, "true": 0.48}, "value": {"0": 0.08, "1": 0.33, "2": 0.59}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:e090dc2104778d4fd274120f", "state_id": "tic_tac_toe:e090dc2104778d4fd274120f", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO O .\n. . X\n. X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.0, "cell_4": 0.0, "cell_5": 0.0, "cell_3": 1.0, "cell_7": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OX.O.X.:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OO...X.X.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.15, "cell_5": 0.33, "cell_3": 0.18, "cell_7": 0.16, "cell_9": 0.18}, "solvable": {"false": 0.79, "true": 0.21}, "value": {"0": 0.25, "1": 0.48, "2": 0.27}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:db5c92533c38a001cad6c895", "state_id": "tic_tac_toe:db5c92533c38a001cad6c895", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X .\n. O O\nX . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_8": "Place X in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_8", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_3": 0.0, "cell_8": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXXO.O.X:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OX..OOX.X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.31, "cell_3": 0.25, "cell_8": 0.44}, "solvable": {"false": 0.52, "true": 0.48}, "value": {"0": 0.19, "1": 0.28, "2": 0.53}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:fc68869a70d1093e310fa7d9", "state_id": "grid_navigation:fc68869a70d1093e310fa7d9", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # . #\n. . . #\n# . G .\n. A # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 1), (1, 3), (2, 0), (3, 0), (3, 2), (3, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [0, 3], [1, 3], [2, 0], [3, 2], [3, 3]], "position": [3, 1], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.72, "west": 0.28}, "solvable": {"false": 0.18000000000000005, "true": 0.82}, "value": {"0": 0.23, "1": 0.57, "2": 0.12, "3": 0.08}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:8cef4e0af9749489b0bb7ea3", "state_id": "tic_tac_toe:8cef4e0af9749489b0bb7ea3", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O O\n. X .\nO X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.3333333333333333, "cell_4": 0.3333333333333333, "cell_6": 0.3333333333333333}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4", "cell_6", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXXOO.X:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOO.X.OX.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.28, "cell_4": 0.27, "cell_9": 0.45}, "solvable": {"false": 0.6, "true": 0.4}, "value": {"0": 0.19, "1": 0.31, "2": 0.5}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:5e9f8ac17489fc2e01d69fc7", "state_id": "tic_tac_toe:5e9f8ac17489fc2e01d69fc7", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\n. O O\nX . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_1": "Place X in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.5, "cell_8": 0.5, "cell_2": 0.0, "cell_3": 0.0, "cell_1": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....OOX.X:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "....OOX.X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.09, "cell_4": 0.32, "cell_8": 0.34, "cell_2": 0.13, "cell_3": 0.12}, "solvable": {"false": 0.6699999999999999, "true": 0.33}, "value": {"0": 0.22, "1": 0.34, "2": 0.44}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:39b5e9de8918f678d76642ed", "state_id": "grid_navigation:39b5e9de8918f678d76642ed", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n# # . G\n# # A .\n# # . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.5, "north": 0.5, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 0), (2, 1), (2, 2), (2, 3), (3, 1), (3, 2), (3, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 0], [1, 1], [2, 0], [2, 1], [3, 0], [3, 1], [3, 3]], "position": [2, 2], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.03, "north": 0.3, "east": 0.67}, "solvable": {"false": 0.28, "true": 0.72}, "value": {"0": 0.21, "1": 0.61, "2": 0.09, "3": 0.09}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:e481e257e5f3f204cbe23f41", "state_id": "tic_tac_toe:e481e257e5f3f204cbe23f41", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . X\n. O O\nX . X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.0, "cell_8": 0.5, "cell_4": 0.5, "cell_2": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..X.O.XOX:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..X.OOX.X", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.17, "cell_4": 0.25, "cell_8": 0.4, "cell_2": 0.18}, "solvable": {"false": 0.88, "true": 0.12}, "value": {"0": 0.5599999999999999, "1": 0.25, "2": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:b2b5814483d536f0663df68a", "state_id": "grid_navigation:b2b5814483d536f0663df68a", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . #\nG . # .\n. # # .\nA . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (1, 2), (2, 1), (2, 2), (3, 0))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 3], [1, 2], [2, 1], [2, 2]], "position": [3, 0], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.73, "east": 0.27}, "solvable": {"false": 0.31000000000000005, "true": 0.69}, "value": {"0": 0.02, "1": 0.44, "2": 0.37, "3": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:7e3e17e8f108b1097a02abe8", "state_id": "tic_tac_toe:7e3e17e8f108b1097a02abe8", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\nO X X\n. O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_1": 0.25, "cell_7": 0.25, "cell_3": 0.25, "cell_2": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_2", "cell_3", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...OXX.OX:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "...OXX.OX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.32, "cell_1": 0.16, "cell_7": 0.37, "cell_2": 0.15}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.54, "1": 0.26, "2": 0.2}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:3e13f12b1c7f80b5eeae383d", "state_id": "tic_tac_toe:3e13f12b1c7f80b5eeae383d", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O .\nX X .\nO O X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place X in row 1, column 3 (cell 3).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_3": 0.3333333333333333, "cell_1": 0.3333333333333333, "cell_6": 0.3333333333333333}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_3", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOXO.XO:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".O.XX.OOX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.25, "cell_1": 0.22, "cell_6": 0.53}, "solvable": {"false": 0.48, "true": 0.52}, "value": {"0": 0.18, "1": 0.29, "2": 0.53}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:1b4c95d674c437fa0de8b45d", "state_id": "tic_tac_toe:1b4c95d674c437fa0de8b45d", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O .\nO X X\n. X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_7": 0.5, "cell_3": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXXXOOX.:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XO.OXX.XO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.49, "cell_3": 0.51}, "solvable": {"false": 0.89, "true": 0.11}, "value": {"0": 0.47, "1": 0.38, "2": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:9026b61505f607c7a8e8f0d0", "state_id": "grid_navigation:9026b61505f607c7a8e8f0d0", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # #\nA . . .\nG . . .\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "east": 0.0, "south": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((2, 3), (3, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3]], "position": [1, 0], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.02, "south": 0.94, "north": 0.04}, "solvable": {"false": 0.06999999999999995, "true": 0.93}, "value": {"0": 0.75, "1": 0.23, "2": 0.02, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:5dcbe8702edfda1d6d0c3ad0", "state_id": "grid_navigation:5dcbe8702edfda1d6d0c3ad0", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # . A\n# G # .\n. # # #\n. # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.5, "south": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 1), (1, 0), (1, 2), (1, 3), (2, 1), (2, 2), (3, 2))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [1, 0], [1, 2], [2, 1], [2, 2], [2, 3], [3, 1]], "position": [0, 3], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.62, "south": 0.38}, "solvable": {"false": 0.54, "true": 0.46}, "value": {"0": 0.2, "1": 0.32, "2": 0.17, "3": 0.31}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:b4dcfa85b661f517286a099e", "state_id": "grid_navigation:b4dcfa85b661f517286a099e", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . #\nA . . .\n# . # .\n. . G .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 1.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 1), (1, 3), (3, 0))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [2, 0], [2, 2]], "position": [1, 0], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.92, "north": 0.08}, "solvable": {"false": 0.13, "true": 0.87}, "value": {"0": 0.01, "1": 0.67, "2": 0.26, "3": 0.06}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:60fd8888a357231572667431", "state_id": "tic_tac_toe:60fd8888a357231572667431", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\n. O X\n. X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_1": "Place X in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_7": 0.0, "cell_3": 0.0, "cell_2": 0.0, "cell_4": 0.0, "cell_1": 1.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....OX.XO:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "....OX.XO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.25, "cell_3": 0.14, "cell_7": 0.28, "cell_2": 0.14, "cell_1": 0.19}, "solvable": {"false": 0.8, "true": 0.2}, "value": {"0": 0.18, "1": 0.44, "2": 0.38}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:34e6e520108ff959e9df4984", "state_id": "tic_tac_toe:34e6e520108ff959e9df4984", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X O\nO . X\n. O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_7": 0.0, "cell_5": 1.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXO.XXXO:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XXOO.X.OX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.62, "cell_7": 0.38}, "solvable": {"false": 0.9, "true": 0.1}, "value": {"0": 0.48, "1": 0.4, "2": 0.12}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:378549ff0a1e958e94443dea", "state_id": "tic_tac_toe:378549ff0a1e958e94443dea", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O .\nO . .\nX O X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place X in row 2, column 2 (cell 5).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_3": "Place X in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 1.0, "cell_6": 0.0, "cell_3": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XO.OXOX:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XO.O..XOX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.2, "cell_5": 0.62, "cell_6": 0.18}, "solvable": {"false": 0.61, "true": 0.39}, "value": {"0": 0.16, "1": 0.38, "2": 0.46}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:6b84d660b5937e5233e3a207", "state_id": "grid_navigation:6b84d660b5937e5233e3a207", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. A . .\nG . # #\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "west": "Move one cell west.", "south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.0, "west": 0.5, "south": 0.5, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((2, 1), (3, 1))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[2, 2], [2, 3]], "position": [1, 1], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.28, "north": 0.01, "east": 0, "west": 0.71}, "solvable": {"false": 0.14, "true": 0.86}, "value": {"0": 0.41, "1": 0.5, "2": 0.07, "3": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:3af976bad7b0ea21902978ea", "state_id": "grid_navigation:3af976bad7b0ea21902978ea", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . A .\n# # # .\n. . . #\n# G . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "west": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (1, 3), (2, 0), (2, 1), (2, 2))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [1, 1], [1, 2], [2, 3], [3, 0]], "position": [0, 2], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.56, "east": 0.44}, "solvable": {"false": 0.32999999999999996, "true": 0.67}, "value": {"0": 0.01, "1": 0.38, "2": 0.44, "3": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:02ec84ed317075102e37e6bf", "state_id": "grid_navigation:02ec84ed317075102e37e6bf", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\nA . # .\nG # . .\n# . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 1.0, "east": 0.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (1, 1), (1, 3), (2, 0), (2, 2))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 2], [2, 1], [3, 0], [3, 2]], "position": [1, 0], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.17, "north": 0.04, "south": 0.79}, "solvable": {"false": 0.44999999999999996, "true": 0.55}, "value": {"0": 0.18, "1": 0.36, "2": 0.23, "3": 0.23}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:7833713cc18f780abd2d1650", "state_id": "grid_navigation:7833713cc18f780abd2d1650", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # .\n. # . .\nA . . .\n. G . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.5, "north": 0.0, "east": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((2, 1), (3, 1), (3, 2))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [1, 1]], "position": [2, 0], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.53, "north": 0.03, "south": 0.44}, "solvable": {"false": 0.09999999999999998, "true": 0.9}, "value": {"0": 0.66, "1": 0.27, "2": 0.06, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:8b2df9bc1b2b3c94277487cf", "state_id": "grid_navigation:8b2df9bc1b2b3c94277487cf", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. . . #\n# . A .\n. # . G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east.", "north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "east": 0.5, "north": 0.0, "south": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (1, 3), (2, 0))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 3], [2, 0], [3, 1]], "position": [2, 2], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.25, "north": 0.14, "west": 0.09, "south": 0.52}, "solvable": {"false": 0.22999999999999998, "true": 0.77}, "value": {"0": 0.26, "1": 0.55, "2": 0.11, "3": 0.08}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:3ca782ca04459c66a1268608", "state_id": "tic_tac_toe:3ca782ca04459c66a1268608", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . O\nX O X\nX X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place O in row 1, column 2 (cell 2).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.5, "cell_9": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.XOXO.XXO:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.OXOXXX.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.9, "cell_2": 0.1}, "solvable": {"false": 0.83, "true": 0.17}, "value": {"0": 0.49, "1": 0.22, "2": 0.29}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:b340bf70f242fac502670cae", "state_id": "tic_tac_toe:b340bf70f242fac502670cae", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . .\nX . X\nX . O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place O in row 1, column 2 (cell 2).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.0, "cell_5": 1.0, "cell_8": 0.0, "cell_3": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OX.XO.X:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O..X.XX.O", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.51, "cell_8": 0.28, "cell_3": 0.08, "cell_2": 0.13}, "solvable": {"false": 0.86, "true": 0.14}, "value": {"0": 0.5, "1": 0.29, "2": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:e32309aea072b47da90b4fe5", "state_id": "grid_navigation:e32309aea072b47da90b4fe5", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# G # .\n# # # .\n# . # A\n. # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.5, "south": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 2), (1, 0), (1, 1), (1, 2), (2, 0), (2, 2), (3, 1), (3, 2))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [1, 0], [1, 1], [1, 2], [2, 0], [2, 2], [3, 1], [3, 2]], "position": [2, 3], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.69, "south": 0.31}, "solvable": {"false": 0.6699999999999999, "true": 0.33}, "value": {"0": 0.02, "1": 0.22, "2": 0.33, "3": 0.43}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:1147d5afa892fb07b604e1ed", "state_id": "grid_navigation:1147d5afa892fb07b604e1ed", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A . .\n. . . .\nG # . .\n# # . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 0.5, "east": 0.0, "south": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (1, 0), (1, 1), (3, 0))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[2, 1], [3, 0], [3, 1], [3, 3]], "position": [0, 1], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.5, "east": 0.08, "south": 0.42}, "solvable": {"false": 0.29000000000000004, "true": 0.71}, "value": {"0": 0.02, "1": 0.47, "2": 0.39, "3": 0.12}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:0d0d711b530f18727bff6d2c", "state_id": "tic_tac_toe:0d0d711b530f18727bff6d2c", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X X\nO X .\nO . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place O in row 3, column 2 (cell 8).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_8": 0.0, "cell_6": 0.0, "cell_1": 1.0, "cell_9": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O.XOXX.:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XXOX.O..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.33999999999999997, "cell_8": 0.25, "cell_1": 0.24, "cell_9": 0.17}, "solvable": {"false": 0.88, "true": 0.12}, "value": {"0": 0.61, "1": 0.23, "2": 0.16}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:bcc9fc47f793b19c289aa09f", "state_id": "tic_tac_toe:bcc9fc47f793b19c289aa09f", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X .\nX X .\nX O O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place O in row 2, column 3 (cell 6).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_6": 0.5, "cell_3": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXXOOXX:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OX.XX.XOO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.58, "cell_6": 0.42}, "solvable": {"false": 0.88, "true": 0.12}, "value": {"0": 0.73, "1": 0.15, "2": 0.12}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:97756398e57503a31e601bd7", "state_id": "tic_tac_toe:97756398e57503a31e601bd7", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . X\nX . .\nO . X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place O in row 1, column 2 (cell 2).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_6", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_2": 0.0, "cell_6": 1.0, "cell_8": 0.0, "cell_5": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.XX..O.X:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.XX..O.X", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.11, "cell_5": 0.58, "cell_2": 0.15, "cell_8": 0.16}, "solvable": {"false": 0.9, "true": 0.1}, "value": {"0": 0.47, "1": 0.36, "2": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:d3ddf96446fd51006a41e188", "state_id": "grid_navigation:d3ddf96446fd51006a41e188", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . A\n# . . G\n. # . .\n# . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "south": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 3), (2, 2), (3, 1), (3, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [2, 1], [3, 0], [3, 2]], "position": [0, 3], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.39, "south": 0.61}, "solvable": {"false": 0.16000000000000003, "true": 0.84}, "value": {"0": 0.64, "1": 0.28, "2": 0.05, "3": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:b11337173d7d14010fa0df50", "state_id": "tic_tac_toe:b11337173d7d14010fa0df50", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O .\n. X O\n. . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place X in row 3, column 2 (cell 8).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_8": 0.25, "cell_1": 0.25, "cell_3": 0.25, "cell_4": 0.0, "cell_7": 0.25}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_3", "cell_7", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....XOXO.:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".O..XO..X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.15, "cell_3": 0.17, "cell_8": 0.16, "cell_1": 0.21, "cell_7": 0.31}, "solvable": {"false": 0.6, "true": 0.4}, "value": {"0": 0.16, "1": 0.27, "2": 0.57}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:0da1f845da8245845db533d6", "state_id": "tic_tac_toe:0da1f845da8245845db533d6", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . X\nX . .\nO O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_1": 0.25, "cell_5": 0.25, "cell_6": 0.25, "cell_2": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_2", "cell_5", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XX..OOX:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..XX..OOX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.23, "cell_2": 0.18, "cell_1": 0.15, "cell_5": 0.44}, "solvable": {"false": 0.86, "true": 0.14}, "value": {"0": 0.47000000000000003, "1": 0.33, "2": 0.2}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:f26f42f35f73186003c02011", "state_id": "grid_navigation:f26f42f35f73186003c02011", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A . .\nG . . .\n. # . .\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.5, "west": 0.5, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 2),)", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[2, 1]], "position": [0, 1], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.1, "east": 0, "west": 0.9}, "solvable": {"false": 0.06000000000000005, "true": 0.94}, "value": {"0": 0.84, "1": 0.15, "2": 0.01, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:d114cec9a0e4e4ede70a76ed", "state_id": "grid_navigation:d114cec9a0e4e4ede70a76ed", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # #\n# . . .\n# G # A\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.5, "north": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (0, 2), (0, 3), (2, 1), (2, 3), (3, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [0, 3], [1, 0], [2, 0], [2, 2]], "position": [2, 3], "goal": [2, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.6, "north": 0.4}, "solvable": {"false": 0.38, "true": 0.62}, "value": {"0": 0.1, "1": 0.43, "2": 0.31, "3": 0.16}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:02e914e4fdb9e4f5b2723ecc", "state_id": "tic_tac_toe:02e914e4fdb9e4f5b2723ecc", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . X\nO O .\nX O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place O in row 1, column 2 (cell 2).", "cell_6": "Place O in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.5, "cell_6": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:X.X.OOXOX:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.XOO.XOX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.61, "cell_2": 0.39}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.49, "1": 0.33, "2": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:575fc6a54a54c3fbff91a1e2", "state_id": "grid_navigation:575fc6a54a54c3fbff91a1e2", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # .\nG # A .\n# . . .\n. # # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"south": 0.5, "east": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 2), (1, 1), (1, 3), (2, 0), (2, 3), (3, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [1, 1], [2, 0], [3, 1], [3, 2], [3, 3]], "position": [1, 2], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.81, "east": 0.19}, "solvable": {"false": 0.6, "true": 0.4}, "value": {"0": 0.06, "1": 0.48, "2": 0.22, "3": 0.24}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:c5ac509cbb6034df446e888a", "state_id": "grid_navigation:c5ac509cbb6034df446e888a", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # # .\n. . . .\n. . A .\n. G . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south.", "north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.0, "south": 0.5, "north": 0.0, "west": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (3, 0), (3, 1), (3, 2))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [0, 2], [3, 3]], "position": [2, 2], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.01, "north": 0.01, "south": 0.4, "west": 0.58}, "solvable": {"false": 0.050000000000000044, "true": 0.95}, "value": {"0": 0.34, "1": 0.61, "2": 0.05, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:6400f8049d76ca96607b5a28", "state_id": "grid_navigation:6400f8049d76ca96607b5a28", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. G . A\n# # . .\n# . . .\n. # . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (1, 1), (2, 0), (3, 1), (3, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [1, 1], [2, 0], [3, 1], [3, 3]], "position": [0, 3], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.22, "west": 0.78}, "solvable": {"false": 0.32999999999999996, "true": 0.67}, "value": {"0": 0.31, "1": 0.47, "2": 0.16, "3": 0.06}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:a77d96343187b2b055300d0d", "state_id": "tic_tac_toe:a77d96343187b2b055300d0d", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X O\nX X .\nO . O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place O in row 3, column 2 (cell 8).", "cell_6": "Place O in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_6", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_8": 0.5, "cell_6": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_6", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.O.XXOXX:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XXOXX.O.O", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.33, "cell_6": 0.67}, "solvable": {"false": 0.84, "true": 0.16}, "value": {"0": 0.6, "1": 0.17, "2": 0.23}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:3bd0a0148ef715b36672d3e5", "state_id": "tic_tac_toe:3bd0a0148ef715b36672d3e5", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X .\nO . X\n. O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.0, "cell_9": 0.0, "cell_1": 0.0, "cell_3": 1.0, "cell_5": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.O.O.X.X.:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".X.O.X.O.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.1, "cell_9": 0.17, "cell_7": 0.09, "cell_3": 0.1, "cell_5": 0.54}, "solvable": {"false": 0.76, "true": 0.24}, "value": {"0": 0.16, "1": 0.5, "2": 0.34}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:da43084e6a1cd9e92e7ae5d2", "state_id": "grid_navigation:da43084e6a1cd9e92e7ae5d2", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . .\n. A . .\n# # . .\nG # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.3333333333333333, "north": 0.3333333333333333, "west": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 3), (1, 0), (1, 1))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [2, 0], [2, 1], [3, 1]], "position": [1, 1], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.1, "east": 0.3, "west": 0.6}, "solvable": {"false": 0.43999999999999995, "true": 0.56}, "value": {"0": 0.02, "1": 0.41, "2": 0.36, "3": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:04090d5dca0c88358efd06b5", "state_id": "grid_navigation:04090d5dca0c88358efd06b5", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # #\n. # A .\nG . . .\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.0, "south": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 2), (2, 3), (3, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [1, 1]], "position": [1, 2], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.08, "south": 0.92}, "solvable": {"false": 0.15000000000000002, "true": 0.85}, "value": {"0": 0.07, "1": 0.75, "2": 0.14, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:2e8f57b6b20af2ea04e2fd47", "state_id": "tic_tac_toe:2e8f57b6b20af2ea04e2fd47", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . .\nX . O\n. . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place X in row 1, column 3 (cell 3).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_3": 0.2, "cell_8": 0.2, "cell_2": 0.2, "cell_7": 0.2, "cell_5": 0.2}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_3", "cell_5", "cell_7", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OO.XX..:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O..X.O..X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.45999999999999996, "cell_3": 0.15, "cell_7": 0.12, "cell_8": 0.15, "cell_2": 0.12}, "solvable": {"false": 0.6599999999999999, "true": 0.34}, "value": {"0": 0.2, "1": 0.39, "2": 0.41}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:a4b967ea242e82f7f9892afd", "state_id": "grid_navigation:a4b967ea242e82f7f9892afd", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # # .\n. . # #\nA . # .\n. # G .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"south": 0.3333333333333333, "north": 0.3333333333333333, "east": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 1), (2, 0), (2, 1), (3, 1), (3, 2), (3, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [0, 2], [1, 2], [1, 3], [2, 2], [3, 1]], "position": [2, 0], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.1, "south": 0.64, "east": 0.26}, "solvable": {"false": 0.41000000000000003, "true": 0.59}, "value": {"0": 0.07, "1": 0.52, "2": 0.22, "3": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:bc246ac7440e6269a943269a", "state_id": "tic_tac_toe:bc246ac7440e6269a943269a", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . .\nO X X\nX . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place O in row 3, column 2 (cell 8).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_8": 0.0, "cell_2": 0.0, "cell_3": 1.0, "cell_9": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXXO..X:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O..OXXX..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.28, "cell_8": 0.32, "cell_3": 0.19, "cell_9": 0.21}, "solvable": {"false": 0.83, "true": 0.17}, "value": {"0": 0.36, "1": 0.33, "2": 0.31}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:1823e4700fa3efc928686204", "state_id": "grid_navigation:1823e4700fa3efc928686204", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # .\n. # . #\n. # G .\n. . A .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "north": 1.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 2), (1, 3), (2, 1), (2, 2), (3, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [1, 1], [1, 3], [2, 1]], "position": [3, 2], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.55, "west": 0.22, "east": 0.23}, "solvable": {"false": 0.44999999999999996, "true": 0.55}, "value": {"0": 0.09, "1": 0.48, "2": 0.25, "3": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:61e79037083a3ae3b19cd044", "state_id": "tic_tac_toe:61e79037083a3ae3b19cd044", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO O .\nX X .\nO . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place X in row 3, column 2 (cell 8).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_6", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_8": 0.0, "cell_3": 0.0, "cell_6": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOX.OXO:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OO.XX.O.X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.37, "cell_3": 0.29, "cell_6": 0.34}, "solvable": {"false": 0.5800000000000001, "true": 0.42}, "value": {"0": 0.25, "1": 0.36, "2": 0.39}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:69ca6f89d7614492a8296bc6", "state_id": "grid_navigation:69ca6f89d7614492a8296bc6", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG . . #\n. # . .\n# # # #\n# . A .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.5, "east": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (0, 3), (1, 1), (1, 2), (2, 2), (3, 0), (3, 2))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 1], [2, 0], [2, 1], [2, 2], [2, 3], [3, 0]], "position": [3, 2], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.49, "east": 0.51}, "solvable": {"false": 0.65, "true": 0.35}, "value": {"0": 0.02, "1": 0.3, "2": 0.33, "3": 0.35000000000000003}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:c19de8d2b1278a80ff92dc4d", "state_id": "tic_tac_toe:c19de8d2b1278a80ff92dc4d", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O .\nO . .\nX X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place X in row 2, column 3 (cell 6).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.0, "cell_3": 0.3333333333333333, "cell_1": 0.0, "cell_5": 0.3333333333333333, "cell_9": 0.3333333333333333}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_5", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...O.X.OX:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".O.O..XX.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.53, "cell_5": 0.19, "cell_3": 0.1, "cell_1": 0.07, "cell_6": 0.11}, "solvable": {"false": 0.38, "true": 0.62}, "value": {"0": 0.08, "1": 0.13, "2": 0.79}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:9ffaf445abc0e5736c02accb", "state_id": "tic_tac_toe:9ffaf445abc0e5736c02accb", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . X\nO . X\nO X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place O in row 2, column 2 (cell 5).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_5": 0.0, "cell_2": 1.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:OOXX..OXX:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.XO.XOXO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.46, "cell_5": 0.54}, "solvable": {"false": 0.86, "true": 0.14}, "value": {"0": 0.43, "1": 0.38, "2": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:84853ff2965085a6aa898e65", "state_id": "grid_navigation:84853ff2965085a6aa898e65", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG . . .\n# . . .\n. . . .\n# # A .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 3), (1, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [3, 0], [3, 1]], "position": [3, 2], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.22, "north": 0.78}, "solvable": {"false": 0.19999999999999996, "true": 0.8}, "value": {"0": 0.02, "1": 0.49, "2": 0.42, "3": 0.07}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:697febb56350023c42c79616", "state_id": "grid_navigation:697febb56350023c42c79616", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n# . . .\n. # . .\nA . # G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.5, "east": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (1, 2), (2, 3), (3, 1))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 0], [2, 1], [3, 2]], "position": [3, 0], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.58, "north": 0.42}, "solvable": {"false": 0.30000000000000004, "true": 0.7}, "value": {"0": 0.03, "1": 0.38, "2": 0.42, "3": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:c4ca519fc1a97b0fa92b135d", "state_id": "grid_navigation:c4ca519fc1a97b0fa92b135d", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nA . . .\n. # G .\n. # . #\n. . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 2), (0, 3), (1, 3), (2, 1), (2, 2))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 1], [2, 1], [2, 3], [3, 2], [3, 3]], "position": [0, 0], "goal": [1, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.73, "south": 0.27}, "solvable": {"false": 0.4, "true": 0.6}, "value": {"0": 0.06, "1": 0.51, "2": 0.25, "3": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:bcc1be04e57ef4c5a057a37c", "state_id": "tic_tac_toe:bcc1be04e57ef4c5a057a37c", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . .\nO . X\nX X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_9": 0.25, "cell_3": 0.25, "cell_5": 0.25, "cell_2": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_3", "cell_5", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OX.O.XX:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O..O.XXX.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.42, "cell_3": 0.17, "cell_5": 0.29, "cell_2": 0.12}, "solvable": {"false": 0.83, "true": 0.17}, "value": {"0": 0.45, "1": 0.23, "2": 0.32}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:038af61143264eec01ae4b92", "state_id": "tic_tac_toe:038af61143264eec01ae4b92", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . X\nO X O\nO X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_9": 0.5, "cell_2": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXXX.OOX:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.XOXOOX.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.82, "cell_2": 0.18}, "solvable": {"false": 0.9, "true": 0.1}, "value": {"0": 0.58, "1": 0.32, "2": 0.1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:1a28116ee6ba6c8a336b87c6", "state_id": "tic_tac_toe:1a28116ee6ba6c8a336b87c6", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . .\nO X O\n. X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place O in row 1, column 3 (cell 3).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_3": 0.25, "cell_7": 0.25, "cell_9": 0.25, "cell_2": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_3", "cell_7", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOXO.X.:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X..OXO.X.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.23, "cell_7": 0.29000000000000004, "cell_3": 0.24, "cell_9": 0.24}, "solvable": {"false": 0.86, "true": 0.14}, "value": {"0": 0.52, "1": 0.33, "2": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:b818845ebb92c160e3e786f6", "state_id": "grid_navigation:b818845ebb92c160e3e786f6", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. G # #\n# . A .\n# . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 2), (0, 3), (2, 1), (2, 3), (3, 1))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 2], [1, 3], [2, 0], [3, 0], [3, 2]], "position": [2, 2], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.17, "west": 0.83}, "solvable": {"false": 0.33999999999999997, "true": 0.66}, "value": {"0": 0.3, "1": 0.48, "2": 0.12, "3": 0.1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:148ab89508b44c72913190e8", "state_id": "tic_tac_toe:148ab89508b44c72913190e8", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . O\n. . X\nO X X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place X in row 1, column 2 (cell 2).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_2": 0.3333333333333333, "cell_4": 0.3333333333333333, "cell_5": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_4", "cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.O..XOXX:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.O..XOXX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.39, "cell_2": 0.3, "cell_4": 0.31}, "solvable": {"false": 0.5800000000000001, "true": 0.42}, "value": {"0": 0.26, "1": 0.29, "2": 0.45}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:da5af3b49ba894bd19ca15dc", "state_id": "tic_tac_toe:da5af3b49ba894bd19ca15dc", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O .\nO . X\nX O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place X in row 1, column 3 (cell 3).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_3": 0.0, "cell_9": 0.0, "cell_5": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXX.O.OX:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XO.O.XXO.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.46, "cell_3": 0.24, "cell_9": 0.3}, "solvable": {"false": 0.6799999999999999, "true": 0.32}, "value": {"0": 0.18, "1": 0.47000000000000003, "2": 0.35}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:2014d23bbab19ccc04946726", "state_id": "tic_tac_toe:2014d23bbab19ccc04946726", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . O\n. O X\n. O X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_7": 0.3333333333333333, "cell_4": 0.3333333333333333, "cell_2": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_4", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOO.XXO:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.O.OX.OX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.26, "cell_4": 0.3, "cell_7": 0.44}, "solvable": {"false": 0.7, "true": 0.3}, "value": {"0": 0.49, "1": 0.2, "2": 0.31}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:2c9768c2ce1e4029e4a8bcad", "state_id": "grid_navigation:2c9768c2ce1e4029e4a8bcad", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG . . .\n. # # #\n# # . #\n. A . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.5, "east": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (1, 1), (1, 2), (2, 1), (3, 1), (3, 2))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 1], [1, 2], [1, 3], [2, 0], [2, 1], [2, 3]], "position": [3, 1], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.34, "east": 0.66}, "solvable": {"false": 0.48, "true": 0.52}, "value": {"0": 0.03, "1": 0.32, "2": 0.42, "3": 0.23}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:06f63fffd6492e4c74df7b9d", "state_id": "grid_navigation:06f63fffd6492e4c74df7b9d", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n# # . .\nA G # .\n. . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 2), (1, 2), (2, 0), (2, 1), (2, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 0], [1, 1], [2, 2], [3, 2]], "position": [2, 0], "goal": [2, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.11, "east": 0.89}, "solvable": {"false": 0.08999999999999997, "true": 0.91}, "value": {"0": 0.97, "1": 0.01, "2": 0.01, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:f316be1cfda62fb555f108f6", "state_id": "grid_navigation:f316be1cfda62fb555f108f6", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A . .\n# . . .\n. # # .\nG . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 2}, "gold_probs": {"action": {"west": 0.0, "east": 0.5, "south": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (1, 1), (2, 1))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [2, 1], [2, 2]], "position": [0, 1], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.07, "south": 0.58, "west": 0.35}, "solvable": {"false": 0.18999999999999995, "true": 0.81}, "value": {"0": 0.01, "1": 0.51, "2": 0.41, "3": 0.07}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:782e6f5e4bef0f98a3830604", "state_id": "grid_navigation:782e6f5e4bef0f98a3830604", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A . .\n. . # .\n# # . #\n. G # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"south": 0.3333333333333333, "east": 0.3333333333333333, "west": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 0), (1, 1), (1, 3), (2, 2))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 2], [2, 0], [2, 1], [2, 3], [3, 2]], "position": [0, 1], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.67, "west": 0.23, "east": 0.1}, "solvable": {"false": 0.38, "true": 0.62}, "value": {"0": 0.02, "1": 0.43, "2": 0.36, "3": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:5b958615e4d17f8ba311c5fe", "state_id": "grid_navigation:5b958615e4d17f8ba311c5fe", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . G\n. . A .\n# . # #\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "north": 0.5, "east": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (1, 2), (3, 2))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[2, 0], [2, 2], [2, 3]], "position": [1, 2], "goal": [0, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.86, "west": 0.06, "east": 0.08}, "solvable": {"false": 0.12, "true": 0.88}, "value": {"0": 0.77, "1": 0.15, "2": 0.05, "3": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:e552c89c6b7dbaf914f4764e", "state_id": "grid_navigation:e552c89c6b7dbaf914f4764e", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. G . .\n# # . #\n# # . .\n. . A .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.0, "north": 1.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (1, 1), (1, 3), (2, 0), (2, 1))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [1, 1], [1, 3], [2, 0], [2, 1]], "position": [3, 2], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.12, "east": 0.1, "north": 0.78}, "solvable": {"false": 0.29000000000000004, "true": 0.71}, "value": {"0": 0.01, "1": 0.44, "2": 0.42, "3": 0.13}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:d677d1e66c59ee1430659556", "state_id": "tic_tac_toe:d677d1e66c59ee1430659556", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . X\nO X X\nO . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place O in row 1, column 2 (cell 2).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_1": "Place O in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.0, "cell_9": 0.5, "cell_8": 0.0, "cell_1": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXXOX..:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..XOXXO..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.16, "cell_8": 0.38, "cell_2": 0.26, "cell_9": 0.2}, "solvable": {"false": 0.86, "true": 0.14}, "value": {"0": 0.68, "1": 0.17, "2": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:d0166c88a4f9de080a24b6c8", "state_id": "grid_navigation:d0166c88a4f9de080a24b6c8", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nA . # .\n. # . .\n# . G .\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"south": 0.5, "east": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((1, 3), (2, 2), (3, 1))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 1], [2, 0]], "position": [0, 0], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.11, "south": 0.89}, "solvable": {"false": 0.21999999999999997, "true": 0.78}, "value": {"0": 0.02, "1": 0.49, "2": 0.41, "3": 0.08}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:a311b3b030af6e443089057f", "state_id": "grid_navigation:a311b3b030af6e443089057f", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . #\n. . . .\n. G # .\n# . . A", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 1.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (1, 2), (3, 1), (3, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 3], [2, 2], [3, 0]], "position": [3, 3], "goal": [2, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.51, "west": 0.49}, "solvable": {"false": 0.19999999999999996, "true": 0.8}, "value": {"0": 0.04, "1": 0.6, "2": 0.28, "3": 0.08}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:139d024f846851f39955759b", "state_id": "grid_navigation:139d024f846851f39955759b", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . .\n. A G #\n# . . .\n# . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south.", "north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "south": 0.0, "north": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (0, 3), (1, 3), (3, 0), (3, 2), (3, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 3], [2, 0], [3, 0], [3, 2], [3, 3]], "position": [1, 1], "goal": [1, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0, "west": 0, "north": 0, "east": 1}, "solvable": {"false": 0.040000000000000036, "true": 0.96}, "value": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:d50040350093b838cdb748ac", "state_id": "tic_tac_toe:d50040350093b838cdb748ac", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . .\n. . X\n. X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_3": 0.0, "cell_7": 0.0, "cell_5": 1.0, "cell_2": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OX..OX.:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O....X.XO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.15, "cell_4": 0.16, "cell_5": 0.36, "cell_2": 0.11, "cell_7": 0.22}, "solvable": {"false": 0.6599999999999999, "true": 0.34}, "value": {"0": 0.17, "1": 0.33, "2": 0.5}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:ac1e3bbcb6b1e4631cc07e12", "state_id": "tic_tac_toe:ac1e3bbcb6b1e4631cc07e12", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . X\n. O X\nO . O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_8": "Place X in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_2": 1.0, "cell_8": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.O.OXX.X:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.X.OXO.O", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.25, "cell_8": 0.43, "cell_2": 0.32}, "solvable": {"false": 0.64, "true": 0.36}, "value": {"0": 0.28, "1": 0.4, "2": 0.32}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:218127f33da48bcec72a8471", "state_id": "grid_navigation:218127f33da48bcec72a8471", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # #\n. G . A\n. # . .\n# . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 2), (0, 3), (2, 1), (3, 0))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [2, 1], [3, 0]], "position": [1, 3], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.89, "south": 0.11}, "solvable": {"false": 0.21999999999999997, "true": 0.78}, "value": {"0": 0.31, "1": 0.5, "2": 0.13, "3": 0.06}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:2a65409dced57a3c577822a1", "state_id": "tic_tac_toe:2a65409dced57a3c577822a1", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O .\nX . X\n. O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.0, "cell_7": 0.0, "cell_3": 0.0, "cell_5": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.O.X.X.OX:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".O.X.X.OX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.13, "cell_7": 0.21, "cell_3": 0.16, "cell_5": 0.5}, "solvable": {"false": 0.91, "true": 0.09}, "value": {"0": 0.54, "1": 0.29, "2": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:f1b87e124e74e79e42437b66", "state_id": "grid_navigation:f1b87e124e74e79e42437b66", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # . .\n# . . .\n. G . .\n# . A #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.5, "north": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 2), (0, 3), (1, 3), (3, 0))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [1, 0], [3, 0], [3, 3]], "position": [3, 2], "goal": [2, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.48, "north": 0.52}, "solvable": {"false": 0.10999999999999999, "true": 0.89}, "value": {"0": 0.24, "1": 0.64, "2": 0.1, "3": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:4e229d3d5f3050dff6b90712", "state_id": "grid_navigation:4e229d3d5f3050dff6b90712", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # . .\n# . . .\n. G . A\n. . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 1.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 2), (0, 3), (1, 3), (3, 0))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [1, 0], [3, 3]], "position": [2, 3], "goal": [2, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.75, "north": 0.25}, "solvable": {"false": 0.06000000000000005, "true": 0.94}, "value": {"0": 0.21, "1": 0.7, "2": 0.08, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:ec97e5ee075a8ef017103d2d", "state_id": "grid_navigation:ec97e5ee075a8ef017103d2d", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG . . .\n. . . #\n. # . .\nA . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 1.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((1, 2), (2, 3), (3, 1))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 3], [2, 1], [3, 2]], "position": [3, 0], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.22, "north": 0.78}, "solvable": {"false": 0.26, "true": 0.74}, "value": {"0": 0.01, "1": 0.39, "2": 0.51, "3": 0.09}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:2c8a5992ef1cd4f971b6cfa9", "state_id": "tic_tac_toe:2c8a5992ef1cd4f971b6cfa9", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X .\nO X X\n. O O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_3": "Place X in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_7", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_1": 0.0, "cell_7": 1.0, "cell_3": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.O.OXXOX.:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".X.OXX.OO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.27, "cell_7": 0.44, "cell_1": 0.29}, "solvable": {"false": 0.53, "true": 0.47}, "value": {"0": 0.25, "1": 0.29, "2": 0.46}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:5e8fce50a960eca94c4dc763", "state_id": "tic_tac_toe:5e8fce50a960eca94c4dc763", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . X\n. O X\nO O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_9": 0.5, "cell_2": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOXO.X.X:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.X.OXOO.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.14, "cell_4": 0.16, "cell_9": 0.7}, "solvable": {"false": 0.47, "true": 0.53}, "value": {"0": 0.27, "1": 0.22, "2": 0.51}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:3f7c383659498fc1c9d5a2a1", "state_id": "tic_tac_toe:3f7c383659498fc1c9d5a2a1", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O .\nX O .\nO X X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place O in row 1, column 3 (cell 3).", "cell_6": "Place O in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_3": 1.0, "cell_6": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOOXXXO:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XO.XO.OXX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.56, "cell_3": 0.44}, "solvable": {"false": 0.81, "true": 0.19}, "value": {"0": 0.44, "1": 0.31, "2": 0.25}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:6f9b48bb617d7ab72b0aff66", "state_id": "tic_tac_toe:6f9b48bb617d7ab72b0aff66", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O .\n. O .\nX X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place X in row 2, column 3 (cell 6).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_3": "Place X in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.0, "cell_4": 1.0, "cell_3": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OOOXX.X:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XO..O.XXO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.39, "cell_6": 0.23, "cell_3": 0.38}, "solvable": {"false": 0.76, "true": 0.24}, "value": {"0": 0.32, "1": 0.45, "2": 0.23}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:d2ebda30df7e8cb4c12e8e85", "state_id": "tic_tac_toe:d2ebda30df7e8cb4c12e8e85", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O X\n. X .\nO . O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_8", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_4": 0.0, "cell_8": 1.0, "cell_6": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.O.X.XOX:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOX.X.O.O", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.2, "cell_8": 0.51, "cell_4": 0.29}, "solvable": {"false": 0.6, "true": 0.4}, "value": {"0": 0.34, "1": 0.34, "2": 0.32}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:d72a38cc75defa6c2cd84d1e", "state_id": "grid_navigation:d72a38cc75defa6c2cd84d1e", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG # . #\n. # # #\nA . . #\n. # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "north": 1.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 3), (1, 1), (1, 2), (1, 3), (2, 3), (3, 1))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 3], [1, 1], [1, 2], [1, 3], [2, 3], [3, 1]], "position": [2, 0], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.53, "east": 0.37, "south": 0.1}, "solvable": {"false": 0.6799999999999999, "true": 0.32}, "value": {"0": 0.04, "1": 0.24, "2": 0.19, "3": 0.53}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:db1fccd90d9f458aa5bc1e61", "state_id": "grid_navigation:db1fccd90d9f458aa5bc1e61", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # .\nG . # #\n. . # .\n. # . A", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 0), (1, 3), (2, 0), (2, 1), (2, 2), (3, 1))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [1, 2], [1, 3], [2, 2], [3, 1]], "position": [3, 3], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.61, "west": 0.39}, "solvable": {"false": 0.37, "true": 0.63}, "value": {"0": 0.01, "1": 0.35, "2": 0.46, "3": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:867c177bceca875373716a27", "state_id": "tic_tac_toe:867c177bceca875373716a27", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O X\n. . X\n. O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_1": "Place X in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.0, "cell_5": 0.5, "cell_9": 0.5, "cell_4": 0.0, "cell_1": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...O.O.XX:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".OX..X.O.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.12, "cell_4": 0.11, "cell_9": 0.3, "cell_7": 0.15, "cell_5": 0.32}, "solvable": {"false": 0.54, "true": 0.46}, "value": {"0": 0.13, "1": 0.37, "2": 0.5}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:e887ae1e4c1e8ae311869c5a", "state_id": "grid_navigation:e887ae1e4c1e8ae311869c5a", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. . . .\n. . A .\nG # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.0, "east": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 2))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[3, 1], [3, 2]], "position": [2, 2], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.21, "east": 0.16, "west": 0.63}, "solvable": {"false": 0.43999999999999995, "true": 0.56}, "value": {"0": 0.04, "1": 0.5, "2": 0.31, "3": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:bd424ae46565f9664baff1f5", "state_id": "grid_navigation:bd424ae46565f9664baff1f5", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n# . . .\n. # . A\n. . G .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.5, "west": 0.5, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 2), (2, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 0], [2, 1]], "position": [2, 3], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.44, "north": 0.11, "south": 0.45}, "solvable": {"false": 0.19999999999999996, "true": 0.8}, "value": {"0": 0.15, "1": 0.67, "2": 0.13, "3": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:af5d9688bd1bc7e79def7f8f", "state_id": "grid_navigation:af5d9688bd1bc7e79def7f8f", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . #\n. # G #\n. . A .\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north.", "west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.0, "north": 1.0, "west": 0.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 1), (2, 1), (3, 0))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 3], [1, 1], [1, 3]], "position": [2, 2], "goal": [1, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.13, "north": 0.73, "south": 0.02, "east": 0.12}, "solvable": {"false": 0.32999999999999996, "true": 0.67}, "value": {"0": 0.34, "1": 0.41, "2": 0.13, "3": 0.12}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:c3f3cf2cc61d8f841f80d953", "state_id": "grid_navigation:c3f3cf2cc61d8f841f80d953", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # . #\n. . # .\nG . . A\n. . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 3), (2, 0), (2, 2), (3, 0), (3, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [0, 3], [1, 2], [3, 2], [3, 3]], "position": [2, 3], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.13, "west": 0.87}, "solvable": {"false": 0.14, "true": 0.86}, "value": {"0": 0.05, "1": 0.66, "2": 0.21, "3": 0.08}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:73e606ba3f8a470629370bb1", "state_id": "tic_tac_toe:73e606ba3f8a470629370bb1", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . O\nX . O\nO . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place X in row 3, column 2 (cell 8).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_8": 0.0, "cell_2": 0.0, "cell_5": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.XO.XX.O:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.OX.OO.X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.44, "cell_2": 0.29, "cell_8": 0.27}, "solvable": {"false": 0.6699999999999999, "true": 0.33}, "value": {"0": 0.22, "1": 0.44, "2": 0.34}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:053b07140234cccc862ba0f4", "state_id": "grid_navigation:053b07140234cccc862ba0f4", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # .\n. . . #\n# G A .\n. . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south.", "north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 1.0, "south": 0.0, "north": 0.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (0, 3), (2, 3), (3, 0), (3, 2))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [1, 3], [2, 0], [3, 3]], "position": [2, 2], "goal": [2, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.01, "south": 0.01, "east": 0, "west": 0.98}, "solvable": {"false": 0.07999999999999996, "true": 0.92}, "value": {"0": 0.86, "1": 0.11, "2": 0.02, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:b5b8be9b202a193b1bbaa6dc", "state_id": "tic_tac_toe:b5b8be9b202a193b1bbaa6dc", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . .\nX . .\nX O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place X in row 1, column 2 (cell 2).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_3": "Place X in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.0, "cell_9": 0.0, "cell_5": 1.0, "cell_6": 0.0, "cell_3": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.....OOXX:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O..X..XO.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.22, "cell_6": 0.23, "cell_3": 0.17, "cell_2": 0.1, "cell_5": 0.28}, "solvable": {"false": 0.6, "true": 0.4}, "value": {"0": 0.13, "1": 0.26, "2": 0.61}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:f947ee03b33f271c3b6cf604", "state_id": "grid_navigation:f947ee03b33f271c3b6cf604", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # .\n. A . #\n# . # .\n. # . G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east.", "south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.25, "east": 0.25, "south": 0.25, "west": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north", "south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 2), (1, 1), (1, 3), (2, 0), (3, 1), (3, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [1, 3], [2, 0], [2, 2], [3, 1]], "position": [1, 1], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.22, "north": 0.12, "east": 0.3, "south": 0.36}, "solvable": {"false": 0.43999999999999995, "true": 0.56}, "value": {"0": 0.01, "1": 0.44, "2": 0.34, "3": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:5d558860062c371c634108c3", "state_id": "tic_tac_toe:5d558860062c371c634108c3", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X X\nX O .\nO O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.5, "cell_6": 0.0, "cell_9": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOOXOX.:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XXXO.OO.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.43, "cell_6": 0.32, "cell_1": 0.25}, "solvable": {"false": 0.48, "true": 0.52}, "value": {"0": 0.26, "1": 0.25, "2": 0.49}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:8a2444e0c4932554d13f8a54", "state_id": "grid_navigation:8a2444e0c4932554d13f8a54", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . .\n# A . .\n. # . .\n. . . G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((1, 2), (2, 3), (3, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 0], [2, 1]], "position": [1, 1], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.95, "north": 0.05}, "solvable": {"false": 0.18999999999999995, "true": 0.81}, "value": {"0": 0.01, "1": 0.61, "2": 0.33, "3": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:fa51fb0307b2708dced2fab5", "state_id": "tic_tac_toe:fa51fb0307b2708dced2fab5", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . X\nX . .\nO X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place O in row 2, column 2 (cell 5).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_5": 0.5, "cell_6": 0.5, "cell_9": 0.0, "cell_2": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XX..OXO:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.XX..OX.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.45, "cell_6": 0.16, "cell_2": 0.15, "cell_9": 0.24}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.4, "1": 0.38, "2": 0.22}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:995b710e0ef485725543631d", "state_id": "grid_navigation:995b710e0ef485725543631d", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. . . A\n# # # .\n. . . G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 1.0, "north": 0.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((1, 1), (1, 2), (1, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[2, 0], [2, 1], [2, 2]], "position": [1, 3], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.04, "west": 0.23, "south": 0.73}, "solvable": {"false": 0.31000000000000005, "true": 0.69}, "value": {"0": 0.12, "1": 0.49, "2": 0.2, "3": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:26004888392d1fd66799de01", "state_id": "grid_navigation:26004888392d1fd66799de01", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # .\n# . # #\n# A G .\n# . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "east": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 2), (1, 2), (1, 3), (3, 0), (3, 1), (3, 2), (3, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [1, 0], [1, 2], [1, 3], [2, 0], [3, 0], [3, 3]], "position": [2, 1], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.02, "south": 0, "east": 0.98}, "solvable": {"false": 0.050000000000000044, "true": 0.95}, "value": {"0": 0.99, "1": 0, "2": 0, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:fbb6f2c2fe25dff7439faab4", "state_id": "tic_tac_toe:fbb6f2c2fe25dff7439faab4", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . X\n. X X\n. O .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_7": 0.25, "cell_2": 0.25, "cell_4": 0.25, "cell_9": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_4", "cell_7", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OOX..XX:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.X.XX.O.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.33, "cell_9": 0.19, "cell_2": 0.12, "cell_4": 0.36}, "solvable": {"false": 0.88, "true": 0.12}, "value": {"0": 0.74, "1": 0.12, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:41b11bf77be4049ad2a90e18", "state_id": "grid_navigation:41b11bf77be4049ad2a90e18", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\n. . # .\nG . . .\nA . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 3), (2, 2))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 2]], "position": [3, 0], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.16, "north": 0.84}, "solvable": {"false": 0.12, "true": 0.88}, "value": {"0": 0.14, "1": 0.65, "2": 0.17, "3": 0.04}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:da9e76cf2dfcc99acdf5d030", "state_id": "tic_tac_toe:da9e76cf2dfcc99acdf5d030", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . .\nO X .\nX O X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place X in row 2, column 3 (cell 6).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_3": "Place X in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.0, "cell_2": 0.0, "cell_3": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O.XOXOX:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O..OX.XOX", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.3, "cell_6": 0.39, "cell_2": 0.31}, "solvable": {"false": 0.62, "true": 0.38}, "value": {"0": 0.22, "1": 0.32, "2": 0.46}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:3c2f5c47a1a3a36bd792a47d", "state_id": "grid_navigation:3c2f5c47a1a3a36bd792a47d", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . #\n. # . .\n. . A .\n. . . G", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east.", "north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "east": 0.5, "north": 0.0, "south": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 3), (2, 2), (3, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 3], [1, 1]], "position": [2, 2], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.05, "north": 0.03, "south": 0.6, "east": 0.32}, "solvable": {"false": 0.13, "true": 0.87}, "value": {"0": 0.38, "1": 0.51, "2": 0.08, "3": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:ea6fa2808788cde84cb5ff9a", "state_id": "tic_tac_toe:ea6fa2808788cde84cb5ff9a", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O .\n. O .\nX . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_8", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_1": 0.0, "cell_8": 1.0, "cell_3": 0.0, "cell_6": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XOO...X:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".O..O.X.X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.19, "cell_3": 0.1, "cell_6": 0.15, "cell_8": 0.45, "cell_1": 0.11}, "solvable": {"false": 0.6599999999999999, "true": 0.34}, "value": {"0": 0.25, "1": 0.39, "2": 0.36}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:160025d1d7d2510204217061", "state_id": "tic_tac_toe:160025d1d7d2510204217061", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X .\nO . .\nX X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_6": "Place O in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_9": 0.25, "cell_3": 0.25, "cell_5": 0.25, "cell_6": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_5", "cell_6", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...X.XOOX:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OX.O..XX.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.18, "cell_5": 0.31, "cell_3": 0.18, "cell_9": 0.33}, "solvable": {"false": 0.85, "true": 0.15}, "value": {"0": 0.57, "1": 0.2, "2": 0.23}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:b42bfbbcd161aa537fa62d88", "state_id": "tic_tac_toe:b42bfbbcd161aa537fa62d88", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O .\n. . X\nX . O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place O in row 2, column 1 (cell 4).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_4": 1.0, "cell_8": 0.0, "cell_5": 0.0, "cell_3": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXX..O.X:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XO...XX.O", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.09, "cell_5": 0.4, "cell_8": 0.36, "cell_4": 0.15}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.57, "1": 0.28, "2": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:b452b75c0844de5f8c6f80ac", "state_id": "grid_navigation:b452b75c0844de5f8c6f80ac", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. G # .\n# . . .\n. # # #\n. A . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.5, "east": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 0), (2, 1), (2, 2), (2, 3), (3, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 0], [2, 1], [2, 2], [2, 3], [3, 3]], "position": [3, 1], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.62, "east": 0.38}, "solvable": {"false": 0.35, "true": 0.65}, "value": {"0": 0.04, "1": 0.56, "2": 0.25, "3": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:745cdb5f4fd414b351c6bd37", "state_id": "grid_navigation:745cdb5f4fd414b351c6bd37", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . #\n. G A .\n. # . #\n. . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west.", "south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "west": 1.0, "south": 0.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 3), (2, 1), (2, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 3], [2, 1], [2, 3]], "position": [1, 2], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0, "west": 0.99, "north": 0.01, "south": 0}, "solvable": {"false": 0.07999999999999996, "true": 0.92}, "value": {"0": 0.9, "1": 0.06, "2": 0.02, "3": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:3880c95827718a7180fc0eab", "state_id": "grid_navigation:3880c95827718a7180fc0eab", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # .\n# . # G\n. # # #\nA . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 0), (1, 1), (1, 2), (1, 3), (2, 0), (2, 2), (3, 1))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [1, 0], [1, 2], [2, 1], [2, 2], [2, 3], [3, 2]], "position": [3, 0], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.46, "north": 0.54}, "solvable": {"false": 0.6, "true": 0.4}, "value": {"0": 0.02, "1": 0.35, "2": 0.21, "3": 0.42}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:611d8afc91f4f055a93f59cb", "state_id": "tic_tac_toe:611d8afc91f4f055a93f59cb", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\nO . X\n. X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place X in row 2, column 2 (cell 5).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_2": "Place X in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_5": 0.25, "cell_7": 0.25, "cell_1": 0.25, "cell_3": 0.0, "cell_2": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_2", "cell_5", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...O.X.XO:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "...O.X.XO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.13, "cell_5": 0.42, "cell_2": 0.15, "cell_1": 0.1, "cell_7": 0.2}, "solvable": {"false": 0.71, "true": 0.29}, "value": {"0": 0.19, "1": 0.44, "2": 0.37}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:7ff768c86f61350581677243", "state_id": "tic_tac_toe:7ff768c86f61350581677243", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . X\nX X .\nO . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place O in row 3, column 2 (cell 8).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_6", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_8": 0.0, "cell_6": 1.0, "cell_9": 0.0, "cell_2": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O.XXX.O:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.XXX.O..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.14, "cell_6": 0.52, "cell_8": 0.17, "cell_2": 0.17}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.61, "1": 0.2, "2": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:6da0eea613ed1a131681abe5", "state_id": "grid_navigation:6da0eea613ed1a131681abe5", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nA . # .\n. # # .\n# . G #\n# . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"south": 0.5, "east": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 1), (1, 2), (1, 3), (2, 2), (3, 0), (3, 1))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 1], [1, 2], [2, 0], [2, 3], [3, 0]], "position": [0, 0], "goal": [2, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.16, "south": 0.84}, "solvable": {"false": 0.33999999999999997, "true": 0.66}, "value": {"0": 0.02, "1": 0.47, "2": 0.36, "3": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:c6b63c07aedd139aa07bd8ac", "state_id": "tic_tac_toe:c6b63c07aedd139aa07bd8ac", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X O\n. X O\n. . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_8", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_8": 1.0, "cell_7": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O.XXXOO:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXO.XO..X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.4, "cell_8": 0.18, "cell_7": 0.42}, "solvable": {"false": 0.52, "true": 0.48}, "value": {"0": 0.12, "1": 0.2, "2": 0.68}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:adce8ed03cf4b326b1618bf1", "state_id": "grid_navigation:adce8ed03cf4b326b1618bf1", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # .\n. G # .\n. # . .\n# # . A", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 2), (1, 2), (2, 1), (3, 0), (3, 1))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 2], [2, 1], [3, 0], [3, 1]], "position": [3, 3], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.76, "west": 0.24}, "solvable": {"false": 0.48, "true": 0.52}, "value": {"0": 0.02, "1": 0.4, "2": 0.36, "3": 0.22}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:0ed70b645299fab10c76794c", "state_id": "tic_tac_toe:0ed70b645299fab10c76794c", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X .\n. . X\n. O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_7": 0.0, "cell_3": 1.0, "cell_4": 0.0, "cell_5": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OO.XXX.:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OX...X.OX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.2, "cell_7": 0.28, "cell_4": 0.17, "cell_5": 0.35}, "solvable": {"false": 0.9299999999999999, "true": 0.07}, "value": {"0": 0.83, "1": 0.12, "2": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:ae147cf3dbb17c1066ef1484", "state_id": "tic_tac_toe:ae147cf3dbb17c1066ef1484", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X X\nO . X\n. O O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place X in row 2, column 2 (cell 5).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_1": "Place X in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 0.0, "cell_7": 0.5, "cell_1": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.O.O.XOXX:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XXO.X.OO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.26, "cell_7": 0.21, "cell_5": 0.53}, "solvable": {"false": 0.56, "true": 0.44}, "value": {"0": 0.2, "1": 0.31, "2": 0.49}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:68121bea062b14f3f0eec84b", "state_id": "tic_tac_toe:68121bea062b14f3f0eec84b", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X O\nX . X\n. . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place O in row 3, column 2 (cell 8).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_8": 0.0, "cell_9": 0.0, "cell_7": 0.0, "cell_5": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...X.XOXO:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXOX.X...", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.67, "cell_8": 0.09, "cell_9": 0.09, "cell_7": 0.15}, "solvable": {"false": 0.91, "true": 0.09}, "value": {"0": 0.48, "1": 0.36, "2": 0.16}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:b0cb8e5691e302d509db67fe", "state_id": "tic_tac_toe:b0cb8e5691e302d509db67fe", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X .\n. X .\nO . O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_8", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_3": 0.0, "cell_1": 0.0, "cell_8": 1.0, "cell_6": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXX...O:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".X..X.O.O", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.09, "cell_4": 0.17, "cell_8": 0.6, "cell_3": 0.04, "cell_6": 0.1}, "solvable": {"false": 0.5, "true": 0.5}, "value": {"0": 0.18, "1": 0.27, "2": 0.55}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:6fdb4a9c0a7d30fdc79dba09", "state_id": "tic_tac_toe:6fdb4a9c0a7d30fdc79dba09", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O X\nO X .\n. . O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_7", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 1.0, "cell_8": 0.0, "cell_6": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OOX.XOX:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOXOX...O", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.49, "cell_8": 0.25, "cell_7": 0.26}, "solvable": {"false": 0.6799999999999999, "true": 0.32}, "value": {"0": 0.18, "1": 0.5, "2": 0.32}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:01ad3a66dba45648dfaff0c9", "state_id": "tic_tac_toe:01ad3a66dba45648dfaff0c9", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O X\n. . .\nX . O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place O in row 2, column 3 (cell 6).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_6": 0.25, "cell_4": 0.25, "cell_8": 0.25, "cell_5": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4", "cell_5", "cell_6", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.X...XOX:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XOX...X.O", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.1, "cell_8": 0.26, "cell_4": 0.19, "cell_5": 0.45}, "solvable": {"false": 0.89, "true": 0.11}, "value": {"0": 0.51, "1": 0.36, "2": 0.13}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:3ceffb21fda249608ebce869", "state_id": "tic_tac_toe:3ceffb21fda249608ebce869", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X .\nO X O\nX . O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place X in row 1, column 3 (cell 3).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_8": "Place X in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_3": 0.5, "cell_1": 0.0, "cell_8": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOXX..OX:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".X.OXOX.O", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.52, "cell_3": 0.18, "cell_1": 0.3}, "solvable": {"false": 0.64, "true": 0.36}, "value": {"0": 0.24, "1": 0.4, "2": 0.36}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:b838da63c4e2c7252a9bee61", "state_id": "grid_navigation:b838da63c4e2c7252a9bee61", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . #\n. # . .\n. . . G\nA . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.5, "north": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 0), (2, 2))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 1], [3, 2]], "position": [3, 0], "goal": [2, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.45, "east": 0.55}, "solvable": {"false": 0.31000000000000005, "true": 0.69}, "value": {"0": 0.02, "1": 0.51, "2": 0.33, "3": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:b75d4e85ac271d7f663863a0", "state_id": "grid_navigation:b75d4e85ac271d7f663863a0", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\nG # # .\n. . . .\n# # A .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 3), (1, 0), (1, 1), (1, 3), (2, 1))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 1], [1, 2], [3, 0], [3, 1]], "position": [3, 2], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.82, "east": 0.18}, "solvable": {"false": 0.41000000000000003, "true": 0.59}, "value": {"0": 0.02, "1": 0.35, "2": 0.48, "3": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:dbf4a88733b06f484e75a01b", "state_id": "tic_tac_toe:dbf4a88733b06f484e75a01b", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\nX O .\n. X O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_1": 1.0, "cell_7": 0.0, "cell_3": 0.0, "cell_2": 0.0, "cell_6": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....OXOX.:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "...XO..XO", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.16, "cell_7": 0.28, "cell_2": 0.12, "cell_6": 0.21, "cell_3": 0.23}, "solvable": {"false": 0.75, "true": 0.25}, "value": {"0": 0.27, "1": 0.41, "2": 0.32}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:b94d42e0342ee675084eb22f", "state_id": "grid_navigation:b94d42e0342ee675084eb22f", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # #\n. . # #\n. A # .\n. G . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "south": 1.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 2), (2, 2), (2, 3), (3, 2), (3, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [0, 3], [1, 2], [1, 3], [2, 2]], "position": [2, 1], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.09, "north": 0.01, "south": 0.9}, "solvable": {"false": 0.09999999999999998, "true": 0.9}, "value": {"0": 0.8099999999999999, "1": 0.11, "2": 0.05, "3": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:868510f89f823159add3aee7", "state_id": "tic_tac_toe:868510f89f823159add3aee7", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X .\nO O .\nX . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_8": "Place X in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_6", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_9": 0.0, "cell_1": 0.0, "cell_6": 1.0, "cell_3": 0.0, "cell_8": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....OXXO.:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".X.OO.X..", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.13, "cell_3": 0.13, "cell_6": 0.34, "cell_8": 0.27, "cell_1": 0.13}, "solvable": {"false": 0.5900000000000001, "true": 0.41}, "value": {"0": 0.2, "1": 0.28, "2": 0.52}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:c59a9f537227a5fe9c83e030", "state_id": "tic_tac_toe:c59a9f537227a5fe9c83e030", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X X\nO . .\n. . O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place X in row 2, column 3 (cell 6).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_1": "Place X in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_6": 0.0, "cell_8": 0.25, "cell_5": 0.25, "cell_7": 0.25, "cell_1": 0.25}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_5", "cell_7", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OO...XX:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XXO....O", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.18, "cell_5": 0.49, "cell_8": 0.12, "cell_6": 0.11, "cell_7": 0.1}, "solvable": {"false": 0.71, "true": 0.29}, "value": {"0": 0.21, "1": 0.43, "2": 0.36}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:8c59b8b90685761a49b9fe09", "state_id": "tic_tac_toe:8c59b8b90685761a49b9fe09", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . O\n. X X\nO . O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place X in row 1, column 2 (cell 2).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_4": "Place X in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.0, "cell_8": 0.5, "cell_4": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.O.XXX.O:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.O.XXO.O", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.33, "cell_8": 0.36, "cell_2": 0.31}, "solvable": {"false": 0.5700000000000001, "true": 0.43}, "value": {"0": 0.33, "1": 0.29, "2": 0.38}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:9efca5bfca4e7689229bc189", "state_id": "grid_navigation:9efca5bfca4e7689229bc189", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # . .\n. A # .\n# G . #\n# . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "south": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 1), (0, 3), (1, 3), (2, 2), (3, 0), (3, 1))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [1, 2], [2, 0], [2, 3], [3, 0], [3, 3]], "position": [1, 1], "goal": [2, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.61, "west": 0.39}, "solvable": {"false": 0.43999999999999995, "true": 0.56}, "value": {"0": 0.33, "1": 0.43, "2": 0.11, "3": 0.13}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:8d057611b581dc9321f450bb", "state_id": "tic_tac_toe:8d057611b581dc9321f450bb", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX X .\n. . .\nO O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place O in row 1, column 3 (cell 3).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_4": "Place O in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_3": 0.25, "cell_5": 0.25, "cell_6": 0.25, "cell_4": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_4", "cell_5", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XX.OX.O:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XX....OOX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.14, "cell_5": 0.23, "cell_3": 0.44, "cell_6": 0.19}, "solvable": {"false": 0.81, "true": 0.19}, "value": {"0": 0.35, "1": 0.39, "2": 0.26}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:d1b1b2f92f972573fbe1ba03", "state_id": "tic_tac_toe:d1b1b2f92f972573fbe1ba03", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . X\nX . O\nO . X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place O in row 1, column 2 (cell 2).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_2": 0.3333333333333333, "cell_8": 0.0, "cell_1": 0.3333333333333333, "cell_5": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_2", "cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XX.OO.X:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..XX.OO.X", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.5, "cell_1": 0.11, "cell_2": 0.18, "cell_8": 0.21}, "solvable": {"false": 0.89, "true": 0.11}, "value": {"0": 0.48, "1": 0.36, "2": 0.16}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:5feedd0d26beffc17e675099", "state_id": "tic_tac_toe:5feedd0d26beffc17e675099", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . X\nX . O\nX O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.0, "cell_2": 0.0, "cell_5": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OXO..XXO:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.XX.OXO.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.5, "cell_5": 0.34, "cell_2": 0.16}, "solvable": {"false": 0.44999999999999996, "true": 0.55}, "value": {"0": 0.09, "1": 0.18, "2": 0.73}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:2124ab6d11220fbba0aa527c", "state_id": "tic_tac_toe:2124ab6d11220fbba0aa527c", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO O .\nX . X\nO . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place X in row 3, column 2 (cell 8).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_8": 0.0, "cell_3": 0.5, "cell_5": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.OOX.XX.O:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OO.X.XO.X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.31, "cell_5": 0.4, "cell_3": 0.29}, "solvable": {"false": 0.48, "true": 0.52}, "value": {"0": 0.13, "1": 0.21, "2": 0.66}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:74337113c7d076ec5dcb0633", "state_id": "grid_navigation:74337113c7d076ec5dcb0633", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . .\n. . A .\n. . . .\n. G # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west.", "east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.5, "west": 0.5, "east": 0.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (3, 0))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [3, 2]], "position": [1, 2], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.21, "east": 0, "south": 0.79, "north": 0}, "solvable": {"false": 0.08999999999999997, "true": 0.91}, "value": {"0": 0.09, "1": 0.8, "2": 0.1, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:ff871bdfb4d2c9fbb6f28381", "state_id": "tic_tac_toe:ff871bdfb4d2c9fbb6f28381", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO X X\n. . O\nO . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place X in row 3, column 2 (cell 8).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_8": 0.0, "cell_4": 1.0, "cell_5": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:O.O..XXOX:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "OXX..OO.X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.31, "cell_4": 0.23, "cell_5": 0.46}, "solvable": {"false": 0.56, "true": 0.44}, "value": {"0": 0.18, "1": 0.32, "2": 0.5}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:72f82f341dd9d0af281471ee", "state_id": "grid_navigation:72f82f341dd9d0af281471ee", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . .\n. # . .\n. . A .\n. . G #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south.", "west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.0, "south": 1.0, "west": 0.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (2, 2))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 1], [3, 3]], "position": [2, 2], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.06, "east": 0.01, "south": 0.91, "north": 0.02}, "solvable": {"false": 0.09999999999999998, "true": 0.9}, "value": {"0": 0.73, "1": 0.19, "2": 0.05, "3": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:b6f51e313889d6860b9daca9", "state_id": "grid_navigation:b6f51e313889d6860b9daca9", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # # #\n. # # .\n. G # A\n# . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.0, "south": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 3), (1, 2), (1, 3), (2, 1), (2, 2), (2, 3), (3, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [0, 2], [0, 3], [1, 1], [1, 2], [2, 2], [3, 0]], "position": [2, 3], "goal": [2, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.34, "south": 0.66}, "solvable": {"false": 0.44999999999999996, "true": 0.55}, "value": {"0": 0.15, "1": 0.52, "2": 0.16, "3": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:ac1e9df8b255b6aa2133de90", "state_id": "grid_navigation:ac1e9df8b255b6aa2133de90", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. G # .\n. . # A\n. . # .\n# . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.5, "south": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 2), (2, 2), (3, 0), (3, 2), (3, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 2], [1, 2], [2, 2], [3, 0], [3, 2], [3, 3]], "position": [1, 3], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.65, "south": 0.35}, "solvable": {"false": 0.53, "true": 0.47}, "value": {"0": 0.06, "1": 0.5, "2": 0.25, "3": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:48dcf3c6d86eb72e315631f2", "state_id": "grid_navigation:48dcf3c6d86eb72e315631f2", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # .\n# . . .\n. G . .\n# . . A", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 0.5, "north": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(1, 1):walls=((0, 0), (0, 2), (1, 3), (2, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [0, 2], [1, 0], [3, 0]], "position": [3, 3], "goal": [2, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.54, "north": 0.46}, "solvable": {"false": 0.13, "true": 0.87}, "value": {"0": 0.07, "1": 0.68, "2": 0.2, "3": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:ec97f6d858b064c7bf4f0638", "state_id": "grid_navigation:ec97f6d858b064c7bf4f0638", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . #\n. . . G\n. . . .\n# A . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.5, "north": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 3), (3, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [3, 0], [3, 3]], "position": [3, 1], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.73, "east": 0.27}, "solvable": {"false": 0.08999999999999997, "true": 0.91}, "value": {"0": 0.04, "1": 0.69, "2": 0.22, "3": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:9c14017cda4be9caca6a97ea", "state_id": "grid_navigation:9c14017cda4be9caca6a97ea", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . .\n# . # #\n# # # .\nG # . A", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.5, "west": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 2), (1, 0), (1, 1), (1, 3), (2, 1), (2, 2), (3, 2))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 0], [1, 2], [1, 3], [2, 0], [2, 1], [2, 2], [3, 1]], "position": [3, 3], "goal": [3, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.5, "north": 0.5}, "solvable": {"false": 0.53, "true": 0.47}, "value": {"0": 0.06, "1": 0.41, "2": 0.28, "3": 0.25}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:5dff648e2dd4cabf35a6e560", "state_id": "tic_tac_toe:5dff648e2dd4cabf35a6e560", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O .\nO X .\nX O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.3333333333333333, "cell_3": 0.3333333333333333, "cell_6": 0.3333333333333333}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_6", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...OXOXOX:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XO.OX.XO.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.28, "cell_3": 0.47, "cell_6": 0.25}, "solvable": {"false": 0.5800000000000001, "true": 0.42}, "value": {"0": 0.13, "1": 0.34, "2": 0.53}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:93ecfd6be35942e6f8a81d8f", "state_id": "grid_navigation:93ecfd6be35942e6f8a81d8f", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . #\n. # A G\n. . . .\n. # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "east": 1.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (2, 1), (2, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 1], [3, 1]], "position": [1, 2], "goal": [1, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.99, "north": 0.01, "south": 0}, "solvable": {"false": 0.050000000000000044, "true": 0.95}, "value": {"0": 0.98, "1": 0.01, "2": 0, "3": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:0ec2f3f2e07f642c72eb796c", "state_id": "grid_navigation:0ec2f3f2e07f642c72eb796c", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# G # #\n# . . #\n. # . .\nA . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 2}, "gold_probs": {"action": {"east": 1.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (0, 2), (0, 3), (1, 0), (1, 3), (2, 1))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 2], [0, 3], [1, 0], [1, 3], [2, 1]], "position": [3, 0], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.39, "north": 0.61}, "solvable": {"false": 0.43999999999999995, "true": 0.56}, "value": {"0": 0.01, "1": 0.33, "2": 0.4, "3": 0.26}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:597d1e3a93dfed13c5cd1169", "state_id": "grid_navigation:597d1e3a93dfed13c5cd1169", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # # .\n. # # #\n# . # G\n. . . A", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 1.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 2), (1, 1), (1, 2), (1, 3), (2, 2), (2, 3), (3, 1), (3, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [0, 1], [0, 2], [1, 1], [1, 2], [1, 3], [2, 0], [2, 2]], "position": [3, 3], "goal": [2, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.32, "north": 0.68}, "solvable": {"false": 0.43999999999999995, "true": 0.56}, "value": {"0": 0.09, "1": 0.55, "2": 0.22, "3": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:00cdde931e35361a8cc4c22d", "state_id": "grid_navigation:00cdde931e35361a8cc4c22d", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. G . .\n. # # #\n. # # #\n. A . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 1.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((1, 1), (1, 2), (1, 3), (2, 1), (2, 2), (2, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[1, 1], [1, 2], [1, 3], [2, 1], [2, 2], [2, 3]], "position": [3, 1], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.51, "east": 0.49}, "solvable": {"false": 0.5800000000000001, "true": 0.42}, "value": {"0": 0.05, "1": 0.36, "2": 0.42, "3": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:ea48f5e6a2895cdc161eb5ac", "state_id": "tic_tac_toe:ea48f5e6a2895cdc161eb5ac", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\nO X O\nX . X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_8": "Place O in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_1": 0.25, "cell_2": 0.25, "cell_3": 0.25, "cell_8": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_2", "cell_3", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...OXOX.X:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "...OXOX.X", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.13, "cell_8": 0.64, "cell_3": 0.06, "cell_2": 0.17}, "solvable": {"false": 0.85, "true": 0.15}, "value": {"0": 0.44, "1": 0.38, "2": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:984734e31d9abe363ee4d7dd", "state_id": "tic_tac_toe:984734e31d9abe363ee4d7dd", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\nX X O\nX . O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place O in row 1, column 2 (cell 2).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.0, "cell_1": 0.0, "cell_8": 0.0, "cell_3": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...OXXO.X:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "...XXOX.O", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.21, "cell_8": 0.58, "cell_3": 0.08, "cell_1": 0.13}, "solvable": {"false": 0.85, "true": 0.15}, "value": {"0": 0.59, "1": 0.25, "2": 0.16}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:37e802d6965f7d8b9154394c", "state_id": "grid_navigation:37e802d6965f7d8b9154394c", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . G\n. . . #\n# A . .\n# # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.5, "north": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (2, 3), (3, 0), (3, 2), (3, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 0], [1, 3], [2, 0], [3, 0], [3, 1]], "position": [2, 1], "goal": [0, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.37, "north": 0.63}, "solvable": {"false": 0.22999999999999998, "true": 0.77}, "value": {"0": 0.03, "1": 0.6, "2": 0.27, "3": 0.1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:9baa63d379f7f5004e45d0a2", "state_id": "grid_navigation:9baa63d379f7f5004e45d0a2", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG # . .\n# # . .\n. # . .\n# . A #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 0):walls=((0, 1), (0, 3), (1, 0), (1, 1), (1, 2), (3, 3))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 1], [1, 0], [1, 1], [2, 1], [3, 0], [3, 3]], "position": [3, 2], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.3, "north": 0.7}, "solvable": {"false": 0.49, "true": 0.51}, "value": {"0": 0.02, "1": 0.28, "2": 0.45, "3": 0.25}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:929a75b8e95617502d09e1fc", "state_id": "tic_tac_toe:929a75b8e95617502d09e1fc", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O X\nX . O\nX . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_1": 0.25, "cell_9": 0.25, "cell_8": 0.25, "cell_5": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_5", "cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XO.XXO.:O", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".OXX.OX..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.16, "cell_5": 0.39, "cell_9": 0.15, "cell_8": 0.3}, "solvable": {"false": 0.88, "true": 0.12}, "value": {"0": 0.67, "1": 0.19, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:c145ac0baed298033779cd9e", "state_id": "grid_navigation:c145ac0baed298033779cd9e", "family_id": "grid_navigation_bfs_v1", "split": "test", "state": "Independent game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . #\n# # . #\n# . # #\n. A G #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.0, "north": 0.0, "east": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:4:goal=(0, 1):walls=((0, 0), (1, 0), (1, 1), (1, 3), (2, 0), (2, 2), (2, 3), (3, 0))", "template_id": "grid_navigation:test:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 4, "walls": [[0, 3], [1, 0], [1, 1], [1, 3], [2, 0], [2, 2], [2, 3], [3, 3]], "position": [3, 1], "goal": [3, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.04, "west": 0.01, "east": 0.95}, "solvable": {"false": 0.08999999999999997, "true": 0.91}, "value": {"0": 0.95, "1": 0.02, "2": 0.01, "3": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:1913794c806dc28a96c8d747", "state_id": "tic_tac_toe:1913794c806dc28a96c8d747", "family_id": "tic_tac_toe_minimax_v1", "split": "test", "state": "Independent game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X O\n. X X\nO O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place X in row 2, column 1 (cell 4).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_1": "Place X in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_4", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.5, "cell_9": 0.5, "cell_1": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_4", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OXXOOX.:X", "template_id": "tic_tac_toe:test:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XO.XXOO.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.26, "cell_1": 0.2, "cell_9": 0.54}, "solvable": {"false": 0.42000000000000004, "true": 0.58}, "value": {"0": 0.21, "1": 0.2, "2": 0.59}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:dcb5c185421bf009cfd2f82d", "state_id": "tic_tac_toe:dcb5c185421bf009cfd2f82d", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . .\nX . .\n. . X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place O in row 2, column 2 (cell 5).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_5": 0.3333333333333333, "cell_6": 0.3333333333333333, "cell_7": 0.0, "cell_8": 0.0, "cell_3": 0.3333333333333333, "cell_2": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_5", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O..XX..:O", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O..X....X", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.5599999999999999, "cell_6": 0.1, "cell_3": 0.08, "cell_8": 0.1, "cell_2": 0.06, "cell_7": 0.1}, "solvable": {"false": 0.92, "true": 0.08}, "value": {"0": 0.35, "1": 0.5, "2": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:d58b7198e4303e9b48b45f4d", "state_id": "tic_tac_toe:d58b7198e4303e9b48b45f4d", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X .\n. . .\n. O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place X in row 1, column 3 (cell 3).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_3": 0.14285714285714285, "cell_4": 0.14285714285714285, "cell_1": 0.14285714285714285, "cell_7": 0.14285714285714285, "cell_9": 0.14285714285714285, "cell_6": 0.14285714285714285, "cell_5": 0.14285714285714285}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_3", "cell_4", "cell_5", "cell_6", "cell_7", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...O.X...:X", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".X.....O.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.01, "cell_9": 0.04, "cell_7": 0.02, "cell_5": 0.77, "cell_3": 0.03, "cell_4": 0.03, "cell_1": 0.1}, "solvable": {"false": 0.86, "true": 0.14}, "value": {"0": 0.07, "1": 0.78, "2": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:b64243ce4972919b3f28eba4", "state_id": "grid_navigation:b64243ce4972919b3f28eba4", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . . . .\n# # . . . .\n. # . . G .\n. . . # A .\n. . . . # #\n. . . . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(1, 2):walls=((0, 4), (1, 4), (2, 3), (4, 1), (4, 2), (5, 1))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[1, 0], [1, 1], [2, 1], [3, 3], [4, 4], [4, 5]], "position": [3, 4], "goal": [2, 4]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.67, "east": 0.33}, "solvable": {"false": 0.19999999999999996, "true": 0.8}, "value": {"0": 0.28, "1": 0.52, "2": 0.14, "3": 0.06}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:73f07800b2e73bb4e19b9608", "state_id": "tic_tac_toe:73f07800b2e73bb4e19b9608", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . .\n. O .\n. . X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place O in row 1, column 3 (cell 3).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_6": "Place O in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_3": 0.0, "cell_4": 0.25, "cell_7": 0.0, "cell_2": 0.25, "cell_8": 0.25, "cell_6": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_4", "cell_6", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..X.O.X..:O", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X...O...X", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.14, "cell_4": 0.17, "cell_7": 0.12, "cell_2": 0.21, "cell_3": 0.18, "cell_8": 0.18}, "solvable": {"false": 0.89, "true": 0.11}, "value": {"0": 0.2, "1": 0.62, "2": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:4fea83dc0d20f8e7b5955c41", "state_id": "grid_navigation:4fea83dc0d20f8e7b5955c41", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # . . # .\n. . . . # #\n# # . # # #\n. # . . # .\n. # # A . #\n. . G # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(0, 2):walls=((0, 3), (1, 1), (1, 2), (1, 5), (2, 1), (2, 4), (3, 0), (3, 1), (3, 3), (3, 4), (3, 5), (4, 4), (4, 5), (5, 0), (5, 1), (5, 4))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 0], [0, 1], [0, 4], [1, 4], [1, 5], [2, 0], [2, 1], [2, 3], [2, 4], [2, 5], [3, 1], [3, 4], [4, 1], [4, 2], [4, 5], [5, 3]], "position": [4, 3], "goal": [5, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.65, "east": 0.35}, "solvable": {"false": 0.37, "true": 0.63}, "value": {"0": 0.03, "1": 0.43, "2": 0.4, "3": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:ad817903b698b943e0b2f75a", "state_id": "tic_tac_toe:ad817903b698b943e0b2f75a", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O .\n. . .\nX . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place X in row 1, column 3 (cell 3).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_4": "Place X in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_3": 0.0, "cell_8": 0.0, "cell_5": 0.3333333333333333, "cell_1": 0.3333333333333333, "cell_9": 0.3333333333333333, "cell_6": 0.0, "cell_4": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_5", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.....OX..:X", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".O....X..", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.05, "cell_6": 0.03, "cell_4": 0.08, "cell_5": 0.5599999999999999, "cell_1": 0.13, "cell_3": 0.06, "cell_9": 0.09}, "solvable": {"false": 0.81, "true": 0.19}, "value": {"0": 0.14, "1": 0.72, "2": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:e8885c9d492ef38456b63970", "state_id": "grid_navigation:e8885c9d492ef38456b63970", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . . . #\n. . . . # #\n. . . . # .\n. # # G # #\n. # . # . #\n# . # . A .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.3333333333333333, "north": 0.3333333333333333, "west": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(2, 2):walls=((0, 1), (0, 2), (0, 4), (0, 5), (1, 2), (1, 3), (1, 4), (2, 1), (3, 0), (3, 2), (4, 1), (4, 2), (5, 0))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 5], [1, 4], [1, 5], [2, 4], [3, 1], [3, 2], [3, 4], [3, 5], [4, 1], [4, 3], [4, 5], [5, 0], [5, 2]], "position": [5, 4], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.1, "west": 0.43, "north": 0.47}, "solvable": {"false": 0.4, "true": 0.6}, "value": {"0": 0.02, "1": 0.28, "2": 0.5, "3": 0.2}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:cfc0de78eb8da6724e5ce9ae", "state_id": "grid_navigation:cfc0de78eb8da6724e5ce9ae", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . . # .\n# # . # . A\n. . # # . .\n. # # # # #\n# # # . G .\n# . . . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": false, "value": 3}, "gold_probs": {"action": {"north": 0.3333333333333333, "west": 0.3333333333333333, "south": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(1, 1):walls=((0, 0), (0, 2), (1, 2), (1, 5), (2, 2), (2, 3), (2, 4), (3, 1), (3, 2), (3, 3), (4, 1), (4, 2), (4, 4), (5, 0), (5, 1), (5, 4))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 4], [1, 0], [1, 1], [1, 3], [2, 2], [2, 3], [3, 1], [3, 2], [3, 3], [3, 4], [3, 5], [4, 0], [4, 1], [4, 2], [5, 0], [5, 5]], "position": [1, 5], "goal": [4, 4]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.43, "west": 0.44, "north": 0.13}, "solvable": {"false": 0.38, "true": 0.62}, "value": {"0": 0.01, "1": 0.15, "2": 0.63, "3": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:b310f7ed8f18eb540323cf97", "state_id": "grid_navigation:b310f7ed8f18eb540323cf97", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG . # # . .\n. . . # . #\n. # A # . #\n# . . . . .\n. # . . # .\n. . # # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(0, 0):walls=((0, 2), (0, 3), (1, 3), (1, 5), (2, 1), (2, 3), (2, 5), (3, 0), (4, 1), (4, 4), (5, 2), (5, 3), (5, 4))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 2], [0, 3], [1, 3], [1, 5], [2, 1], [2, 3], [2, 5], [3, 0], [4, 1], [4, 4], [5, 2], [5, 3], [5, 4]], "position": [2, 2], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.57, "north": 0.43}, "solvable": {"false": 0.43999999999999995, "true": 0.56}, "value": {"0": 0.02, "1": 0.19, "2": 0.5, "3": 0.29}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:0cab4e4b399b6e43969fdd7c", "state_id": "grid_navigation:0cab4e4b399b6e43969fdd7c", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . . . .\n. G . . . #\n. . . . . .\n. . # . A .\n# # # # # .\n. . . . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"west": 0.5, "east": 0.0, "north": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(1, 1):walls=((0, 4), (1, 4), (2, 3), (2, 4), (3, 4), (4, 4), (4, 5), (5, 1))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[1, 5], [3, 2], [4, 0], [4, 1], [4, 2], [4, 3], [4, 4], [5, 4]], "position": [3, 4], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.48, "north": 0.5, "east": 0.02}, "solvable": {"false": 0.18999999999999995, "true": 0.81}, "value": {"0": 0.01, "1": 0.37, "2": 0.55, "3": 0.07}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:b52f95a1576dd752b79470e7", "state_id": "tic_tac_toe:b52f95a1576dd752b79470e7", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . X\n. . .\n. . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_7": "Place O in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_1": 0.0, "cell_2": 0.0, "cell_9": 0.0, "cell_4": 0.0, "cell_8": 0.0, "cell_6": 0.0, "cell_5": 1.0, "cell_7": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:........X:O", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..X......", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.06, "cell_9": 0.02, "cell_7": 0.01, "cell_2": 0.08, "cell_5": 0.8, "cell_6": 0.01, "cell_8": 0, "cell_4": 0.02}, "solvable": {"false": 0.94, "true": 0.06}, "value": {"0": 0.07, "1": 0.86, "2": 0.07}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:cb4ea4879539ad09758feb2d", "state_id": "tic_tac_toe:cb4ea4879539ad09758feb2d", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\n. X .\nX . O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place O in row 3, column 2 (cell 8).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_8": 0.0, "cell_1": 0.0, "cell_6": 0.0, "cell_4": 0.0, "cell_3": 1.0, "cell_2": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....X.O.X:O", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "....X.X.O", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.25, "cell_3": 0.11, "cell_8": 0.25, "cell_2": 0.15, "cell_6": 0.11, "cell_1": 0.13}, "solvable": {"false": 0.88, "true": 0.12}, "value": {"0": 0.37, "1": 0.45, "2": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:b7fa4f04f176f3d0224baf71", "state_id": "grid_navigation:b7fa4f04f176f3d0224baf71", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . . . .\n# # A . . .\n. . . . . .\n. # # # . .\n. # . . # G\n. . . . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 2}, "gold_probs": {"action": {"south": 0.5, "east": 0.5, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(0, 1):walls=((0, 0), (1, 0), (1, 1), (2, 2), (3, 2), (4, 1), (4, 2), (4, 4), (5, 4))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[1, 0], [1, 1], [3, 1], [3, 2], [3, 3], [4, 1], [4, 4], [5, 4], [5, 5]], "position": [1, 2], "goal": [4, 5]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.05, "south": 0.19, "east": 0.76}, "solvable": {"false": 0.16000000000000003, "true": 0.84}, "value": {"0": 0, "1": 0.14, "2": 0.77, "3": 0.09}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:1cf87aa565e78abfdbbb5789", "state_id": "tic_tac_toe:1cf87aa565e78abfdbbb5789", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X O\nX . .\n. . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place O in row 2, column 2 (cell 5).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_7": "Place O in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 0.0, "cell_9": 1.0, "cell_1": 0.0, "cell_6": 0.0, "cell_8": 0.0, "cell_7": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.....XOX.:O", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XOX.....", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.08, "cell_9": 0.05, "cell_8": 0.04, "cell_7": 0.07, "cell_5": 0.55, "cell_1": 0.21}, "solvable": {"false": 0.9299999999999999, "true": 0.07}, "value": {"0": 0.38, "1": 0.49, "2": 0.13}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:f4a9e25ee59cefb6367a2d8c", "state_id": "grid_navigation:f4a9e25ee59cefb6367a2d8c", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . . # .\n. . . . . .\n. . . # . .\n. . . G . .\n. . A . . .\n. . . . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north.", "west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "north": 0.5, "west": 0.0, "east": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(2, 2):walls=((1, 5), (2, 3), (4, 5))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 1], [0, 4], [2, 3]], "position": [4, 2], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.06, "north": 0.71, "east": 0.19, "south": 0.04}, "solvable": {"false": 0.06999999999999995, "true": 0.93}, "value": {"0": 0.32, "1": 0.55, "2": 0.11, "3": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:708bf723ae48ad58fc279f68", "state_id": "tic_tac_toe:708bf723ae48ad58fc279f68", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\nX . .\nO X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place O in row 2, column 2 (cell 5).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_5": 0.3333333333333333, "cell_9": 0.0, "cell_2": 0.3333333333333333, "cell_6": 0.3333333333333333, "cell_1": 0.0, "cell_3": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_5", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.....X.XO:O", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "...X..OX.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.16, "cell_2": 0.16, "cell_3": 0.11, "cell_9": 0.14, "cell_5": 0.29, "cell_1": 0.14}, "solvable": {"false": 0.88, "true": 0.12}, "value": {"0": 0.32, "1": 0.49, "2": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:5d070d9fff0da6ab3faaee10", "state_id": "tic_tac_toe:5d070d9fff0da6ab3faaee10", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\n. O X\n. X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_4": "Place O in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_9": 0.3333333333333333, "cell_7": 0.3333333333333333, "cell_3": 0.3333333333333333, "cell_1": 0.0, "cell_2": 0.0, "cell_4": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_7", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....OX.X.:O", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "....OX.X.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.14, "cell_4": 0.26, "cell_3": 0.11, "cell_1": 0.12, "cell_9": 0.19, "cell_7": 0.18}, "solvable": {"false": 0.89, "true": 0.11}, "value": {"0": 0.36, "1": 0.47, "2": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:8fe26bff4e345e97d2c6ed21", "state_id": "tic_tac_toe:8fe26bff4e345e97d2c6ed21", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . .\n. . .\nO X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place O in row 1, column 3 (cell 3).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_3": 0.0, "cell_9": 0.0, "cell_4": 0.0, "cell_6": 0.0, "cell_5": 0.5, "cell_2": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.....XX.O:O", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.....OX.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.08, "cell_5": 0.63, "cell_9": 0.08, "cell_2": 0.11, "cell_6": 0.05, "cell_3": 0.05}, "solvable": {"false": 0.9, "true": 0.1}, "value": {"0": 0.26, "1": 0.56, "2": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:2f9555006f0d9ac63232c5b5", "state_id": "grid_navigation:2f9555006f0d9ac63232c5b5", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . . . .\nA G . # # #\n. . # . # .\n# . . # . .\n. . # . . .\n# . . . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "east": 1.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(1, 1):walls=((0, 3), (0, 5), (2, 2), (2, 4), (3, 1), (3, 3), (4, 1), (4, 2), (5, 1))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[1, 3], [1, 4], [1, 5], [2, 2], [2, 4], [3, 0], [3, 3], [4, 2], [5, 0]], "position": [1, 0], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.04, "south": 0.04, "east": 0.92}, "solvable": {"false": 0.040000000000000036, "true": 0.96}, "value": {"0": 0.99, "1": 0.01, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:8cb24143a20ede6627b385ef", "state_id": "tic_tac_toe:8cb24143a20ede6627b385ef", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\n. . .\nX X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place O in row 1, column 3 (cell 3).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_3": 0.5, "cell_4": 0.0, "cell_2": 0.0, "cell_1": 0.0, "cell_6": 0.5, "cell_5": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:......OXX:O", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "......XXO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.13, "cell_6": 0.19, "cell_5": 0.26, "cell_1": 0.16, "cell_2": 0.16, "cell_3": 0.1}, "solvable": {"false": 0.84, "true": 0.16}, "value": {"0": 0.28, "1": 0.44, "2": 0.28}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:a35f5600dad52263eb66689e", "state_id": "tic_tac_toe:a35f5600dad52263eb66689e", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . .\n. . .\nX . O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place O in row 2, column 3 (cell 6).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_6": 0.16666666666666666, "cell_3": 0.16666666666666666, "cell_2": 0.16666666666666666, "cell_8": 0.16666666666666666, "cell_4": 0.16666666666666666, "cell_5": 0.16666666666666666}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_3", "cell_4", "cell_5", "cell_6", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O...X.X:O", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X.....X.O", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.1, "cell_6": 0.05, "cell_4": 0.15, "cell_5": 0.39, "cell_8": 0.14, "cell_3": 0.17}, "solvable": {"false": 0.85, "true": 0.15}, "value": {"0": 0.27, "1": 0.51, "2": 0.22}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:24ae00770505b295eb9f09e9", "state_id": "grid_navigation:24ae00770505b295eb9f09e9", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A # . . .\n# . . . # .\n# . . . . .\n. . . . . .\n. . . . # G\n. # . . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 2}, "gold_probs": {"action": {"south": 1.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(0, 1):walls=((1, 0), (1, 1), (1, 4), (3, 5), (4, 0), (5, 3), (5, 4))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 2], [1, 0], [1, 4], [2, 0], [4, 4], [5, 1], [5, 4]], "position": [0, 1], "goal": [4, 5]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.22, "south": 0.78}, "solvable": {"false": 0.16000000000000003, "true": 0.84}, "value": {"0": 0, "1": 0.08, "2": 0.83, "3": 0.09}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:ae086a9b64a87144f7e78a4e", "state_id": "grid_navigation:ae086a9b64a87144f7e78a4e", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . # . .\n. . # . A #\n. # # # . #\n. # . . . #\n. . # # G #\n. # . . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 1.0, "north": 0.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(1, 1):walls=((0, 1), (0, 2), (0, 3), (0, 4), (2, 1), (2, 3), (2, 5), (3, 1), (3, 3), (3, 4), (4, 0), (4, 2), (4, 3))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 3], [1, 2], [1, 5], [2, 1], [2, 2], [2, 3], [2, 5], [3, 1], [3, 5], [4, 2], [4, 3], [4, 5], [5, 1]], "position": [1, 4], "goal": [4, 4]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.25, "south": 0.64, "north": 0.11}, "solvable": {"false": 0.38, "true": 0.62}, "value": {"0": 0.02, "1": 0.39, "2": 0.43, "3": 0.16}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:395ad71bc879ea3eb0affef2", "state_id": "tic_tac_toe:395ad71bc879ea3eb0affef2", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O X\n. . .\nX . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_1": "Place O in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_9": 0.0, "cell_5": 1.0, "cell_8": 0.0, "cell_4": 0.0, "cell_6": 0.0, "cell_1": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..X...XO.:O", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".OX...X..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.08, "cell_9": 0.08, "cell_1": 0.22, "cell_4": 0.16, "cell_6": 0.06, "cell_5": 0.4}, "solvable": {"false": 0.91, "true": 0.09}, "value": {"0": 0.29, "1": 0.55, "2": 0.16}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:4777d39faf253b1f92160451", "state_id": "grid_navigation:4777d39faf253b1f92160451", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nA . G # . #\n. . . . . .\n# # . . . .\n# . . . . .\n. . . # . .\n# . . . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"east": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(0, 2):walls=((0, 3), (0, 5), (2, 0), (2, 1), (3, 0), (4, 3), (5, 0))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 3], [0, 5], [2, 0], [2, 1], [3, 0], [4, 3], [5, 0]], "position": [0, 0], "goal": [0, 2]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.55, "south": 0.45}, "solvable": {"false": 0.09999999999999998, "true": 0.9}, "value": {"0": 0.9, "1": 0.05, "2": 0.03, "3": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:c890b3682838e2a024b2630f", "state_id": "grid_navigation:c890b3682838e2a024b2630f", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # # G .\n# . # . # #\n. . . # A .\n. . # . . #\n. . . # . .\n. . . . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "south": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(0, 1):walls=((0, 2), (0, 3), (1, 0), (1, 1), (1, 3), (1, 5), (2, 2), (3, 0), (3, 3), (4, 2))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 2], [0, 3], [1, 0], [1, 2], [1, 4], [1, 5], [2, 3], [3, 2], [3, 5], [4, 3]], "position": [2, 4], "goal": [0, 4]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.47, "south": 0.53}, "solvable": {"false": 0.5700000000000001, "true": 0.43}, "value": {"0": 0.02, "1": 0.24, "2": 0.5, "3": 0.24}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:02c93e432a963cfe9327e0e0", "state_id": "tic_tac_toe:02c93e432a963cfe9327e0e0", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\n. . .\n. . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place X in row 3, column 3 (cell 9).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_8": "Place X in row 3, column 2 (cell 8)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_9": 0.1111111111111111, "cell_7": 0.1111111111111111, "cell_5": 0.1111111111111111, "cell_3": 0.1111111111111111, "cell_2": 0.1111111111111111, "cell_1": 0.1111111111111111, "cell_4": 0.1111111111111111, "cell_6": 0.1111111111111111, "cell_8": 0.1111111111111111}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_2", "cell_3", "cell_4", "cell_5", "cell_6", "cell_7", "cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.........:X", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".........", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0, "cell_2": 0, "cell_4": 0, "cell_1": 0.09999999999999999, "cell_5": 0.9, "cell_9": 0, "cell_7": 0, "cell_3": 0, "cell_6": 0}, "solvable": {"false": 0.94, "true": 0.06}, "value": {"0": 0, "1": 0.99, "2": 0.01}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:dcb85d190a5352283dec4767", "state_id": "grid_navigation:dcb85d190a5352283dec4767", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . A . # #\n. G # # . .\n# . # . . .\n. # # # . .\n. # . # # #\n# # # . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 1.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(1, 1):walls=((0, 2), (0, 5), (1, 3), (1, 4), (1, 5), (2, 1), (2, 2), (2, 3), (2, 5), (3, 1), (3, 3), (3, 4), (4, 0), (4, 4), (4, 5), (5, 0), (5, 4), (5, 5))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 4], [0, 5], [1, 2], [1, 3], [2, 0], [2, 2], [3, 1], [3, 2], [3, 3], [4, 1], [4, 3], [4, 4], [4, 5], [5, 0], [5, 1], [5, 2], [5, 4], [5, 5]], "position": [0, 2], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.05, "west": 0.95}, "solvable": {"false": 0.44999999999999996, "true": 0.55}, "value": {"0": 0.15, "1": 0.38, "2": 0.25, "3": 0.22}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:2bbef993a0fa79de8aaf97b2", "state_id": "grid_navigation:2bbef993a0fa79de8aaf97b2", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . A . #\n# . . . . #\n. . . . # .\n. # . # . .\n. . . # . .\n. . # G # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"south": 0.3333333333333333, "west": 0.3333333333333333, "east": 0.3333333333333333}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(0, 2):walls=((0, 1), (0, 3), (1, 2), (2, 2), (2, 4), (3, 1), (4, 0), (4, 5), (5, 0), (5, 4))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 1], [0, 5], [1, 0], [1, 5], [2, 4], [3, 1], [3, 3], [4, 3], [5, 2], [5, 4]], "position": [0, 3], "goal": [5, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.51, "east": 0.06, "west": 0.43}, "solvable": {"false": 0.37, "true": 0.63}, "value": {"0": 0, "1": 0.1, "2": 0.66, "3": 0.24}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:f497dcb41e2981c9585c118d", "state_id": "tic_tac_toe:f497dcb41e2981c9585c118d", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X .\n. O .\n. . X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place O in row 3, column 2 (cell 8).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_8": 0.0, "cell_6": 0.25, "cell_4": 0.25, "cell_1": 0.25, "cell_7": 0.0, "cell_3": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_3", "cell_4", "cell_6"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....OXX..:O", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".X..O...X", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.21000000000000002, "cell_3": 0.18, "cell_7": 0.19, "cell_6": 0.1, "cell_8": 0.15, "cell_1": 0.17}, "solvable": {"false": 0.89, "true": 0.11}, "value": {"0": 0.32, "1": 0.5, "2": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:5912ac89e3a2344f758b6dd7", "state_id": "grid_navigation:5912ac89e3a2344f758b6dd7", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . # # #\n. # . . # .\n. . . . . .\n. # . # # .\n. . # # . A\nG # . . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 2}, "gold_probs": {"action": {"north": 1.0, "south": 0.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(0, 0):walls=((0, 1), (1, 2), (1, 3), (2, 1), (2, 3), (2, 4), (4, 1), (4, 4), (5, 0), (5, 3), (5, 4), (5, 5))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 0], [0, 3], [0, 4], [0, 5], [1, 1], [1, 4], [3, 1], [3, 3], [3, 4], [4, 2], [4, 3], [5, 1]], "position": [4, 5], "goal": [5, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.16, "south": 0.38, "west": 0.46}, "solvable": {"false": 0.18999999999999995, "true": 0.81}, "value": {"0": 0.01, "1": 0.13, "2": 0.76, "3": 0.1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:61a108db41202094a939d629", "state_id": "tic_tac_toe:61a108db41202094a939d629", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\n. O .\n. X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place X in row 1, column 1 (cell 1).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_6": "Place X in row 2, column 3 (cell 6)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_1": 0.16666666666666666, "cell_7": 0.16666666666666666, "cell_9": 0.16666666666666666, "cell_4": 0.16666666666666666, "cell_3": 0.16666666666666666, "cell_2": 0.0, "cell_6": 0.16666666666666666}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_3", "cell_4", "cell_6", "cell_7", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....O..X.:X", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "....O..X.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.05, "cell_6": 0.1, "cell_2": 0.1, "cell_9": 0.14, "cell_4": 0.29, "cell_1": 0.11, "cell_7": 0.21}, "solvable": {"false": 0.84, "true": 0.16}, "value": {"0": 0.19, "1": 0.62, "2": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:057ac73cb5f4966bbf84d760", "state_id": "grid_navigation:057ac73cb5f4966bbf84d760", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . # . .\n# # # . . #\nA . # . G #\n. . # . . #\n. . . # . #\n# . . # . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"south": 0.5, "east": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(1, 2):walls=((0, 1), (0, 2), (0, 3), (0, 4), (0, 5), (2, 0), (2, 4), (2, 5), (3, 1), (3, 2), (3, 3), (4, 1), (5, 1), (5, 5))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 3], [1, 0], [1, 1], [1, 2], [1, 5], [2, 2], [2, 5], [3, 2], [3, 5], [4, 3], [4, 5], [5, 0], [5, 3], [5, 5]], "position": [2, 0], "goal": [2, 4]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.32, "east": 0.68}, "solvable": {"false": 0.41000000000000003, "true": 0.59}, "value": {"0": 0.01, "1": 0.27, "2": 0.5, "3": 0.22}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:be07290cbc2d38a7dc04ddf2", "state_id": "tic_tac_toe:be07290cbc2d38a7dc04ddf2", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . .\n. O .\n. . .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_6": "Place X in row 2, column 3 (cell 6).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_7": "Place X in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_6": 0.14285714285714285, "cell_3": 0.14285714285714285, "cell_4": 0.14285714285714285, "cell_8": 0.14285714285714285, "cell_2": 0.14285714285714285, "cell_9": 0.14285714285714285, "cell_7": 0.14285714285714285}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_3", "cell_4", "cell_6", "cell_7", "cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....O...X:X", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X...O....", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_8": 0.04, "cell_6": 0.04, "cell_4": 0.33, "cell_2": 0.23, "cell_3": 0.12, "cell_7": 0.06, "cell_9": 0.18}, "solvable": {"false": 0.86, "true": 0.14}, "value": {"0": 0.06, "1": 0.77, "2": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:6945fa53f5b4490eb955a4b7", "state_id": "grid_navigation:6945fa53f5b4490eb955a4b7", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . . # .\n. # . . . #\nG . # . . .\n. . # . . .\n# . . A . #\n. . . . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north.", "east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.0, "north": 0.0, "east": 0.0, "west": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(0, 2):walls=((0, 4), (1, 1), (2, 2), (2, 3), (4, 0), (5, 1), (5, 4))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 4], [1, 1], [1, 5], [2, 2], [3, 2], [4, 0], [4, 5]], "position": [4, 3], "goal": [2, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.68, "north": 0.19, "south": 0.09, "east": 0.04}, "solvable": {"false": 0.4, "true": 0.6}, "value": {"0": 0.01, "1": 0.13, "2": 0.7, "3": 0.16}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:95ae7116d8174e7a9a3e7dfd", "state_id": "tic_tac_toe:95ae7116d8174e7a9a3e7dfd", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\nX . .\nX O .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place O in row 2, column 2 (cell 5).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_5": 0.16666666666666666, "cell_1": 0.16666666666666666, "cell_6": 0.16666666666666666, "cell_3": 0.16666666666666666, "cell_2": 0.16666666666666666, "cell_9": 0.16666666666666666}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_2", "cell_3", "cell_5", "cell_6", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.....O.XX:O", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "...X..XO.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.31, "cell_5": 0.14, "cell_3": 0.08, "cell_2": 0.19, "cell_6": 0.11, "cell_9": 0.17}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.42, "1": 0.35, "2": 0.23}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:aa08ddc8349012176c2a5125", "state_id": "tic_tac_toe:aa08ddc8349012176c2a5125", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X X\nO . .\n. . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_1": "Place O in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.0, "cell_8": 0.0, "cell_6": 0.0, "cell_9": 0.0, "cell_5": 0.0, "cell_1": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.....OXX.:O", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".XXO.....", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.05, "cell_5": 0.54, "cell_8": 0.05, "cell_7": 0.03, "cell_9": 0.03, "cell_1": 0.3}, "solvable": {"false": 0.9, "true": 0.1}, "value": {"0": 0.45, "1": 0.37, "2": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:0b749193141a0132ab52c509", "state_id": "tic_tac_toe:0b749193141a0132ab52c509", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\nX . X\nO . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_1": "Place O in row 1, column 1 (cell 1).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_1": 0.0, "cell_5": 1.0, "cell_8": 0.0, "cell_2": 0.0, "cell_3": 0.0, "cell_9": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...X.X..O:O", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "...X.XO..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.56, "cell_3": 0.08, "cell_9": 0.06, "cell_2": 0.17, "cell_8": 0.08, "cell_1": 0.05}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.33, "1": 0.46, "2": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:bb9720fc36c5c3b754fea447", "state_id": "tic_tac_toe:bb9720fc36c5c3b754fea447", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . O\n. . .\n. X .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place X in row 3, column 1 (cell 7).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_9": "Place X in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_7": 0.0, "cell_1": 0.0, "cell_2": 0.0, "cell_6": 0.0, "cell_4": 0.0, "cell_5": 0.0, "cell_9": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.....XO..:X", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..O....X.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.05, "cell_4": 0.07, "cell_6": 0.04, "cell_1": 0.06, "cell_2": 0.08, "cell_5": 0.63, "cell_7": 0.07}, "solvable": {"false": 0.88, "true": 0.12}, "value": {"0": 0.21, "1": 0.62, "2": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:1457668cce763643f6f6c642", "state_id": "grid_navigation:1457668cce763643f6f6c642", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # . . .\n. . . . G .\n# # . . . .\n. . . . # .\n. . A . # .\n. # . # # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "south": "Move one cell south.", "north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 1}, "gold_probs": {"action": {"east": 0.5, "south": 0.0, "north": 0.5, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(1, 1):walls=((0, 3), (2, 4), (2, 5), (3, 1), (4, 1), (5, 0), (5, 1), (5, 2), (5, 4))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 2], [2, 0], [2, 1], [3, 4], [4, 4], [5, 1], [5, 3], [5, 4], [5, 5]], "position": [4, 2], "goal": [1, 4]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.11, "south": 0.02, "north": 0.64, "east": 0.23}, "solvable": {"false": 0.15000000000000002, "true": 0.85}, "value": {"0": 0.01, "1": 0.41, "2": 0.51, "3": 0.07}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:f5fb131bfbdb3a86b0d4735b", "state_id": "grid_navigation:f5fb131bfbdb3a86b0d4735b", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . # . .\n. . . . # #\n. . A . . .\n. G . . . .\n. # . # . #\n. . . . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east.", "south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.5, "east": 0.0, "south": 0.5, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(1, 2):walls=((1, 1), (3, 1), (3, 5), (4, 0), (4, 4), (5, 1), (5, 4))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 3], [1, 4], [1, 5], [4, 1], [4, 3], [4, 5], [5, 4]], "position": [2, 2], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.01, "north": 0.02, "south": 0.43, "west": 0.54}, "solvable": {"false": 0.08999999999999997, "true": 0.91}, "value": {"0": 0.15, "1": 0.68, "2": 0.15, "3": 0.02}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:c840dac0f68b5f9fdc91105e", "state_id": "grid_navigation:c840dac0f68b5f9fdc91105e", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG # . # . .\n# # # . # .\n. . . . # #\n# A . . # #\n# # . . . .\n. . # . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(0, 0):walls=((0, 1), (0, 3), (0, 4), (1, 0), (1, 1), (1, 4), (2, 1), (2, 5), (3, 0), (4, 1), (4, 2), (4, 3), (5, 2), (5, 3))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 1], [0, 3], [1, 0], [1, 1], [1, 2], [1, 4], [2, 4], [2, 5], [3, 0], [3, 4], [3, 5], [4, 0], [4, 1], [5, 2]], "position": [3, 1], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.34, "east": 0.66}, "solvable": {"false": 0.4, "true": 0.6}, "value": {"0": 0.01, "1": 0.17, "2": 0.58, "3": 0.24}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:3117b6eee7c7a5ca10bdba5a", "state_id": "grid_navigation:3117b6eee7c7a5ca10bdba5a", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# # . . . .\n. . # . # .\n# . # . . G\n. . . . A .\n. . . . . #\n. . . . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east.", "south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 0}, "gold_probs": {"action": {"north": 0.5, "east": 0.5, "south": 0.0, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(0, 2):walls=((0, 4), (0, 5), (1, 1), (3, 1), (3, 2), (4, 0), (5, 0), (5, 2))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 0], [0, 1], [1, 2], [1, 4], [2, 0], [2, 2], [4, 5], [5, 5]], "position": [3, 4], "goal": [2, 5]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.4, "south": 0.01, "north": 0.47, "west": 0.12}, "solvable": {"false": 0.20999999999999996, "true": 0.79}, "value": {"0": 0.43, "1": 0.4, "2": 0.12, "3": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:362c7a27f8d373cbefa1e823", "state_id": "grid_navigation:362c7a27f8d373cbefa1e823", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . . # #\n. # A # . .\n. # . # . #\n. # # . . G\n. . . . . #\n. . . . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 2}, "gold_probs": {"action": {"south": 0.0, "north": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(0, 2):walls=((0, 1), (0, 3), (0, 5), (1, 5), (2, 3), (2, 4), (3, 2), (4, 2), (4, 3), (4, 4))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 4], [0, 5], [1, 1], [1, 3], [2, 1], [2, 3], [2, 5], [3, 1], [3, 2], [4, 5]], "position": [1, 2], "goal": [3, 5]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.83, "north": 0.17}, "solvable": {"false": 0.35, "true": 0.65}, "value": {"0": 0.01, "1": 0.29, "2": 0.53, "3": 0.17}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:f58089787f755e7ff2b4ecaf", "state_id": "tic_tac_toe:f58089787f755e7ff2b4ecaf", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\n. X O\n. X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_7": "Place O in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_9": 0.16666666666666666, "cell_2": 0.16666666666666666, "cell_4": 0.16666666666666666, "cell_1": 0.16666666666666666, "cell_3": 0.16666666666666666, "cell_7": 0.16666666666666666}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_2", "cell_3", "cell_4", "cell_7", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....XO.X.:O", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "....XO.X.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.29, "cell_2": 0.18, "cell_1": 0.07, "cell_7": 0.21, "cell_3": 0.08, "cell_9": 0.17}, "solvable": {"false": 0.85, "true": 0.15}, "value": {"0": 0.41, "1": 0.38, "2": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:83fcf7e10ce0f3c88ff6df6c", "state_id": "grid_navigation:83fcf7e10ce0f3c88ff6df6c", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\nG . # . . .\n# # . . # .\n. . . # . .\n# . . . . #\n# A . # # .\n. # . . # #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"east": 0.5, "north": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(0, 0):walls=((0, 1), (0, 3), (0, 4), (1, 1), (1, 5), (2, 0), (3, 2), (3, 4), (4, 1), (4, 4), (4, 5), (5, 3), (5, 5))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 2], [1, 0], [1, 1], [1, 4], [2, 3], [3, 0], [3, 5], [4, 0], [4, 3], [4, 4], [5, 1], [5, 4], [5, 5]], "position": [4, 1], "goal": [0, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.41, "north": 0.59}, "solvable": {"false": 0.4, "true": 0.6}, "value": {"0": 0.01, "1": 0.1, "2": 0.67, "3": 0.22}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:cecf40257953ffd163471808", "state_id": "grid_navigation:cecf40257953ffd163471808", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . # . A\n# G . . . .\n. # . # . .\n# . . . # #\n. . . . . .\n# . . . . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.5, "west": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(1, 1):walls=((0, 0), (0, 1), (0, 3), (0, 5), (1, 2), (3, 0), (3, 2), (4, 3), (5, 3), (5, 5))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 0], [0, 3], [1, 0], [2, 1], [2, 3], [3, 0], [3, 4], [3, 5], [5, 0], [5, 5]], "position": [0, 5], "goal": [1, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.79, "south": 0.21}, "solvable": {"false": 0.18000000000000005, "true": 0.82}, "value": {"0": 0.04, "1": 0.55, "2": 0.33, "3": 0.07}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:16389abd1cc13d39ec630250", "state_id": "grid_navigation:16389abd1cc13d39ec630250", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . . # #\n. . # . G .\nA # . . . #\n. . # . # .\n# # # . . .\n. . # # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 2}, "gold_probs": {"action": {"north": 1.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(1, 1):walls=((0, 0), (0, 1), (1, 3), (2, 0), (2, 4), (3, 1), (3, 3), (4, 3), (4, 4), (4, 5), (5, 1), (5, 2), (5, 3))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 4], [0, 5], [1, 2], [2, 1], [2, 5], [3, 2], [3, 4], [4, 0], [4, 1], [4, 2], [5, 2], [5, 3], [5, 4]], "position": [2, 0], "goal": [1, 4]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.6, "south": 0.4}, "solvable": {"false": 0.28, "true": 0.72}, "value": {"0": 0.01, "1": 0.29, "2": 0.59, "3": 0.11}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:4aa9d6c561dbe656420062f6", "state_id": "grid_navigation:4aa9d6c561dbe656420062f6", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # A . . .\n. . . # . #\n# . . . . .\n. . . . # .\n. . # . # G\n. . . . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 2}, "gold_probs": {"action": {"south": 0.5, "east": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(0, 1):walls=((0, 4), (1, 1), (1, 2), (2, 4), (3, 1), (4, 5), (5, 3))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 1], [1, 3], [1, 5], [2, 0], [3, 4], [4, 2], [4, 4]], "position": [0, 2], "goal": [4, 5]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.45, "east": 0.55}, "solvable": {"false": 0.15000000000000002, "true": 0.85}, "value": {"0": 0, "1": 0.07, "2": 0.84, "3": 0.09}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:1b0a3549d5849d4b343f4afa", "state_id": "tic_tac_toe:1b0a3549d5849d4b343f4afa", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\nX . .\n. O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place X in row 1, column 2 (cell 2).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_5": "Place X in row 2, column 2 (cell 5).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_3": "Place X in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.0, "cell_1": 0.0, "cell_9": 0.0, "cell_5": 0.5, "cell_7": 0.5, "cell_6": 0.0, "cell_3": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5", "cell_7"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.....O.X.:X", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "...X...O.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.55, "cell_9": 0.07, "cell_2": 0.1, "cell_6": 0.07, "cell_1": 0.09, "cell_3": 0.03, "cell_7": 0.09}, "solvable": {"false": 0.81, "true": 0.19}, "value": {"0": 0.12, "1": 0.68, "2": 0.2}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:6e9b1e56d10107aad2fa3dc9", "state_id": "grid_navigation:6e9b1e56d10107aad2fa3dc9", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # . # #\n. . . . # #\nA . . . . .\n. # . . . #\n. . . . # .\n. G . # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.0, "east": 0.0, "south": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(0, 1):walls=((0, 3), (1, 4), (2, 1), (2, 5), (4, 4), (4, 5), (5, 2), (5, 4), (5, 5))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 2], [0, 4], [0, 5], [1, 4], [1, 5], [3, 1], [3, 5], [4, 4], [5, 3]], "position": [2, 0], "goal": [5, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.6, "north": 0.08, "east": 0.32}, "solvable": {"false": 0.15000000000000002, "true": 0.85}, "value": {"0": 0.01, "1": 0.37, "2": 0.56, "3": 0.06}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:7c245d425c50515ef9114c43", "state_id": "tic_tac_toe:7c245d425c50515ef9114c43", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . .\n. . X\nX . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_8": "Place O in row 3, column 2 (cell 8).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_8": 0.0, "cell_4": 0.0, "cell_5": 0.0, "cell_2": 0.0, "cell_3": 1.0, "cell_9": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..OX....X:O", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O....XX..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.48, "cell_4": 0.16, "cell_2": 0.07, "cell_8": 0.08, "cell_3": 0.12, "cell_9": 0.09}, "solvable": {"false": 0.89, "true": 0.11}, "value": {"0": 0.3, "1": 0.48, "2": 0.22}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:ac3a21bce79e6f152bf43ab3", "state_id": "tic_tac_toe:ac3a21bce79e6f152bf43ab3", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. O .\n. . .\nX X .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place O in row 1, column 3 (cell 3).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_9", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_3": 0.0, "cell_4": 0.0, "cell_9": 1.0, "cell_6": 0.0, "cell_1": 0.0, "cell_5": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...O.X..X:O", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".O....XX.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.04, "cell_9": 0.5, "cell_5": 0.13, "cell_4": 0.04, "cell_3": 0.06, "cell_6": 0.23}, "solvable": {"false": 0.86, "true": 0.14}, "value": {"0": 0.34, "1": 0.34, "2": 0.32}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:3ea333b957354303d1ef30e2", "state_id": "tic_tac_toe:3ea333b957354303d1ef30e2", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\n. . O\n. . X\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_5": "Place X in row 2, column 2 (cell 5).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_1": "Place X in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_5": 0.3333333333333333, "cell_2": 0.0, "cell_7": 0.3333333333333333, "cell_8": 0.3333333333333333, "cell_3": 0.0, "cell_4": 0.0, "cell_1": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5", "cell_7", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.......OX:X", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".....O..X", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.11, "cell_3": 0.08, "cell_7": 0.11, "cell_5": 0.45, "cell_1": 0.06, "cell_8": 0.15, "cell_2": 0.04}, "solvable": {"false": 0.8200000000000001, "true": 0.18}, "value": {"0": 0.21, "1": 0.54, "2": 0.25}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:071ff12bfafc707a1a04b23c", "state_id": "grid_navigation:071ff12bfafc707a1a04b23c", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # # . # .\nG . . . # .\n. . . A . .\n. # . . . .\n. . . . . .\n. . . . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north.", "east": "Move one cell east.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 0.0, "north": 0.5, "east": 0.0, "west": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(0, 1):walls=((1, 0), (1, 3), (2, 0), (4, 0), (4, 1))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 1], [0, 2], [0, 4], [1, 4], [3, 1]], "position": [2, 3], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.8, "south": 0.02, "east": 0.01, "north": 0.16}, "solvable": {"false": 0.09999999999999998, "true": 0.9}, "value": {"0": 0.01, "1": 0.48, "2": 0.46, "3": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:c8a9972c66e868921bc56671", "state_id": "tic_tac_toe:c8a9972c66e868921bc56671", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . .\n. . X\n. . X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place O in row 2, column 1 (cell 4).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_4": 0.0, "cell_5": 0.0, "cell_8": 0.0, "cell_2": 0.0, "cell_7": 0.0, "cell_3": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..O...XX.:O", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O....X..X", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.1, "cell_7": 0.17, "cell_3": 0.21, "cell_8": 0.14, "cell_2": 0.05, "cell_5": 0.33}, "solvable": {"false": 0.89, "true": 0.11}, "value": {"0": 0.6, "1": 0.21, "2": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:1095961446795b6eb37687e6", "state_id": "tic_tac_toe:1095961446795b6eb37687e6", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\nX O X\n. . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_9": "Place O in row 3, column 3 (cell 9).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_7": "Place O in row 3, column 1 (cell 7)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_9": 0.16666666666666666, "cell_2": 0.16666666666666666, "cell_3": 0.16666666666666666, "cell_1": 0.16666666666666666, "cell_8": 0.16666666666666666, "cell_7": 0.16666666666666666}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_2", "cell_3", "cell_7", "cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...XOX...:O", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "...XOX...", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.35000000000000003, "cell_8": 0.11, "cell_7": 0.09, "cell_3": 0.11, "cell_1": 0.3, "cell_9": 0.04}, "solvable": {"false": 0.9299999999999999, "true": 0.07}, "value": {"0": 0.26, "1": 0.63, "2": 0.11}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:83c9d3f5004eae17ee2eef8f", "state_id": "tic_tac_toe:83c9d3f5004eae17ee2eef8f", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\n. . X\n. . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place O in row 2, column 1 (cell 4).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_4": 0.25, "cell_7": 0.0, "cell_8": 0.0, "cell_3": 0.25, "cell_5": 0.25, "cell_1": 0.0, "cell_2": 0.0, "cell_9": 0.25}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_4", "cell_5", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.......X.:O", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".....X...", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_7": 0.01, "cell_2": 0.01, "cell_5": 0.85, "cell_4": 0.05, "cell_8": 0.02, "cell_3": 0.02, "cell_9": 0.01, "cell_1": 0.03}, "solvable": {"false": 0.9299999999999999, "true": 0.07}, "value": {"0": 0.13, "1": 0.79, "2": 0.08}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:a63d9dc350d50a33d3d5c766", "state_id": "tic_tac_toe:a63d9dc350d50a33d3d5c766", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . .\n. . O\nX . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place O in row 1, column 3 (cell 3).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_4": "Place O in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_3": 0.16666666666666666, "cell_8": 0.16666666666666666, "cell_5": 0.16666666666666666, "cell_2": 0.16666666666666666, "cell_9": 0.16666666666666666, "cell_4": 0.16666666666666666}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_3", "cell_4", "cell_5", "cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XO....X:O", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X....OX..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_2": 0.16, "cell_5": 0.45, "cell_3": 0.1, "cell_8": 0.06, "cell_9": 0.07, "cell_4": 0.16}, "solvable": {"false": 0.89, "true": 0.11}, "value": {"0": 0.29, "1": 0.51, "2": 0.2}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:cc6c86186701e77ee74885a1", "state_id": "tic_tac_toe:cc6c86186701e77ee74885a1", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . X\nX . .\n. O .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place O in row 1, column 2 (cell 2).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_9": "Place O in row 3, column 3 (cell 9)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_2": 0.0, "cell_1": 0.0, "cell_5": 1.0, "cell_7": 0.0, "cell_6": 0.0, "cell_9": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..XO...X.:O", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "..XX...O.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_1": 0.09, "cell_9": 0.07, "cell_7": 0.08, "cell_2": 0.14, "cell_6": 0.08, "cell_5": 0.54}, "solvable": {"false": 0.91, "true": 0.09}, "value": {"0": 0.35, "1": 0.5, "2": 0.15}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:76dc3aba0734d63e9551443a", "state_id": "grid_navigation:76dc3aba0734d63e9551443a", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . . . .\n. # . . . .\n. . # . # .\n. . . . . A\n. . . . # .\n. . # G . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "west": "Move one cell west.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.0, "west": 0.5, "south": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(0, 2):walls=((0, 3), (1, 1), (3, 1), (3, 3), (4, 4), (5, 4))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 1], [1, 1], [2, 2], [2, 4], [4, 4], [5, 2]], "position": [3, 5], "goal": [5, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.26, "west": 0.64, "north": 0.1}, "solvable": {"false": 0.24, "true": 0.76}, "value": {"0": 0.01, "1": 0.43, "2": 0.49, "3": 0.07}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:80abee25d3d63a1365b6ab47", "state_id": "tic_tac_toe:80abee25d3d63a1365b6ab47", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX O .\n. . .\nX . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place O in row 2, column 1 (cell 4).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_3", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_4": 0.16666666666666666, "cell_5": 0.16666666666666666, "cell_9": 0.16666666666666666, "cell_8": 0.16666666666666666, "cell_6": 0.16666666666666666, "cell_3": 0.16666666666666666}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_3", "cell_4", "cell_5", "cell_6", "cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.....OX.X:O", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "XO....X..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_5": 0.35000000000000003, "cell_3": 0.25, "cell_6": 0.04, "cell_4": 0.18, "cell_8": 0.13, "cell_9": 0.05}, "solvable": {"false": 0.9, "true": 0.1}, "value": {"0": 0.4, "1": 0.44, "2": 0.16}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:e063b6afaadb839260af1c2b", "state_id": "tic_tac_toe:e063b6afaadb839260af1c2b", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\n. X .\n. O .\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place X in row 1, column 3 (cell 3).", "cell_9": "Place X in row 3, column 3 (cell 9).", "cell_2": "Place X in row 1, column 2 (cell 2).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_7": "Place X in row 3, column 1 (cell 7).", "cell_4": "Place X in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_3": 0.16666666666666666, "cell_9": 0.16666666666666666, "cell_2": 0.0, "cell_1": 0.16666666666666666, "cell_6": 0.16666666666666666, "cell_7": 0.16666666666666666, "cell_4": 0.16666666666666666}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_3", "cell_4", "cell_6", "cell_7", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....X..O.:X", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "....X..O.", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.12, "cell_2": 0.12, "cell_4": 0.2, "cell_3": 0.05, "cell_7": 0.17, "cell_9": 0.14, "cell_1": 0.2}, "solvable": {"false": 0.8200000000000001, "true": 0.18}, "value": {"0": 0.1, "1": 0.67, "2": 0.23}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:d1ddbacf49a08c7b3f7e49ca", "state_id": "tic_tac_toe:d1ddbacf49a08c7b3f7e49ca", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nX . .\n. X .\n. O .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place O in row 1, column 2 (cell 2).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_3": "Place O in row 1, column 3 (cell 3)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_2": 0.16666666666666666, "cell_9": 0.16666666666666666, "cell_6": 0.16666666666666666, "cell_4": 0.16666666666666666, "cell_7": 0.16666666666666666, "cell_3": 0.16666666666666666}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_3", "cell_4", "cell_6", "cell_7", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....XOX..:O", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "X...X..O.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_3": 0.17, "cell_6": 0.09, "cell_9": 0.17, "cell_2": 0.16, "cell_4": 0.22, "cell_7": 0.19}, "solvable": {"false": 0.9, "true": 0.1}, "value": {"0": 0.2, "1": 0.62, "2": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:66a945d83032f72ea65b677c", "state_id": "tic_tac_toe:66a945d83032f72ea65b677c", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\n. X .\n. X O\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_7": "Place O in row 3, column 1 (cell 7).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_4": "Place O in row 2, column 1 (cell 4)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_7": 0.0, "cell_3": 0.0, "cell_1": 0.0, "cell_2": 1.0, "cell_6": 0.0, "cell_4": 0.0}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:....X..XO:O", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "....X..XO", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.25, "cell_2": 0.21, "cell_6": 0.16, "cell_3": 0.06, "cell_1": 0.12, "cell_7": 0.2}, "solvable": {"false": 0.87, "true": 0.13}, "value": {"0": 0.41, "1": 0.41, "2": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:7c617ddc7735281b4c3d2ea0", "state_id": "grid_navigation:7c617ddc7735281b4c3d2ea0", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . . . #\n. A . # . G\n. # . . . #\n. . . . . #\n. # . # # #\n# # # . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"east": "Move one cell east.", "north": "Move one cell north.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": true, "value": 2}, "gold_probs": {"action": {"east": 0.5, "north": 0.5, "west": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(0, 1):walls=((0, 0), (0, 2), (0, 3), (0, 4), (1, 4), (1, 5), (2, 1), (2, 4), (3, 5), (4, 2), (4, 4), (4, 5), (5, 5))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 5], [1, 3], [2, 1], [2, 5], [3, 5], [4, 1], [4, 3], [4, 4], [4, 5], [5, 0], [5, 1], [5, 2], [5, 4]], "position": [1, 1], "goal": [1, 5]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"west": 0.07, "east": 0.68, "north": 0.25}, "solvable": {"false": 0.26, "true": 0.74}, "value": {"0": 0.02, "1": 0.53, "2": 0.26, "3": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:5a5f8313a833bcf9248f7ab5", "state_id": "grid_navigation:5a5f8313a833bcf9248f7ab5", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . . # G .\n# . . . . .\n. # . # . .\n. # . . # #\n. # . . A .\n. . . . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "west", "solvable": true, "value": 2}, "gold_probs": {"action": {"west": 1.0, "east": 0.0, "south": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(0, 1):walls=((0, 2), (0, 5), (1, 5), (2, 2), (2, 4), (3, 0), (3, 1), (3, 4), (4, 4))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 0], [0, 3], [1, 0], [2, 1], [2, 3], [3, 1], [3, 4], [3, 5], [4, 1]], "position": [4, 4], "goal": [0, 4]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.11, "south": 0.24, "west": 0.65}, "solvable": {"false": 0.15000000000000002, "true": 0.85}, "value": {"0": 0.01, "1": 0.16, "2": 0.75, "3": 0.08}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:d390186663453eee21e1bfc1", "state_id": "tic_tac_toe:d390186663453eee21e1bfc1", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\n. . .\nX . O\nTo move: X.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place X in row 1, column 2 (cell 2).", "cell_8": "Place X in row 3, column 2 (cell 8).", "cell_4": "Place X in row 2, column 1 (cell 4).", "cell_1": "Place X in row 1, column 1 (cell 1).", "cell_3": "Place X in row 1, column 3 (cell 3).", "cell_6": "Place X in row 2, column 3 (cell 6).", "cell_5": "Place X in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_2": 0.0, "cell_8": 0.0, "cell_4": 0.3333333333333333, "cell_1": 0.3333333333333333, "cell_3": 0.3333333333333333, "cell_6": 0.0, "cell_5": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_3", "cell_4"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:......O.X:X", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "......X.O", "player": "X"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.1, "cell_4": 0.05, "cell_5": 0.36, "cell_8": 0.32, "cell_3": 0.12, "cell_2": 0.02, "cell_1": 0.03}, "solvable": {"false": 0.81, "true": 0.19}, "value": {"0": 0.19, "1": 0.58, "2": 0.23}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:afcb9bc281264112555c2ee6", "state_id": "tic_tac_toe:afcb9bc281264112555c2ee6", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\nX . X\n. O .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place O in row 1, column 3 (cell 3).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_2": "Place O in row 1, column 2 (cell 2)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_5", "solvable": true, "value": 2}, "gold_probs": {"action": {"cell_3": 0.0, "cell_9": 0.0, "cell_5": 1.0, "cell_7": 0.0, "cell_1": 0.0, "cell_2": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 0.0, "2": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_5"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...X.X.O.:O", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "...X.X.O.", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_9": 0.05, "cell_5": 0.48, "cell_3": 0.07, "cell_2": 0.24, "cell_1": 0.09, "cell_7": 0.07}, "solvable": {"false": 0.89, "true": 0.11}, "value": {"0": 0.4, "1": 0.42, "2": 0.18}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:727c75df5414a1d30691545a", "state_id": "grid_navigation:727c75df5414a1d30691545a", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . # # # .\n. . . . . .\n# G # # . #\n# . # . # .\n. # # . A .\n. # . . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.5, "east": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(1, 2):walls=((0, 2), (0, 3), (1, 4), (1, 5), (2, 0), (2, 2), (2, 3), (2, 4), (3, 0), (3, 2), (4, 0), (4, 3), (4, 5), (5, 2))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 2], [0, 3], [0, 4], [2, 0], [2, 2], [2, 3], [2, 5], [3, 0], [3, 2], [3, 4], [4, 1], [4, 2], [5, 1], [5, 4]], "position": [4, 4], "goal": [2, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.09, "west": 0.91}, "solvable": {"false": 0.42000000000000004, "true": 0.58}, "value": {"0": 0.01, "1": 0.16, "2": 0.62, "3": 0.21}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:41b861ded73f64cd717d3361", "state_id": "grid_navigation:41b861ded73f64cd717d3361", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . # # #\n. . . # # .\n. . . . # .\n# . . G . #\n. . . # # #\n. # # . A .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "east", "solvable": false, "value": 3}, "gold_probs": {"action": {"west": 0.5, "east": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 1.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["east", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(2, 2):walls=((0, 1), (0, 2), (0, 5), (1, 1), (1, 3), (1, 4), (1, 5), (2, 1), (2, 4), (2, 5), (3, 0), (4, 0), (5, 2))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 3], [0, 4], [0, 5], [1, 3], [1, 4], [2, 4], [3, 0], [3, 5], [4, 3], [4, 4], [4, 5], [5, 1], [5, 2]], "position": [5, 4], "goal": [3, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.09, "west": 0.91}, "solvable": {"false": 0.27, "true": 0.73}, "value": {"0": 0.02, "1": 0.31, "2": 0.53, "3": 0.14}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:e341ce521fb25ae4d178ae63", "state_id": "tic_tac_toe:e341ce521fb25ae4d178ae63", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. . .\n. . X\nO . X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_4": "Place O in row 2, column 1 (cell 4).", "cell_1": "Place O in row 1, column 1 (cell 1).", "cell_2": "Place O in row 1, column 2 (cell 2).", "cell_3": "Place O in row 1, column 3 (cell 3).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_4": 0.16666666666666666, "cell_1": 0.16666666666666666, "cell_2": 0.16666666666666666, "cell_3": 0.16666666666666666, "cell_8": 0.16666666666666666, "cell_5": 0.16666666666666666}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_2", "cell_3", "cell_4", "cell_5", "cell_8"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:.....XO.X:O", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".....XO.X", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.1, "cell_3": 0.21, "cell_5": 0.35, "cell_2": 0.08, "cell_8": 0.23, "cell_1": 0.03}, "solvable": {"false": 0.86, "true": 0.14}, "value": {"0": 0.4, "1": 0.41, "2": 0.19}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:726196357eac03bc78123ec0", "state_id": "grid_navigation:726196357eac03bc78123ec0", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n# . # . . .\n# . . # # .\n. . . # # A\n. # # . . .\n# . # # . G\n# # # # . #", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 1.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(0, 1):walls=((0, 0), (1, 3), (1, 4), (2, 0), (2, 1), (2, 3), (2, 4), (3, 0), (3, 1), (3, 2), (3, 5), (4, 0), (4, 2), (5, 0), (5, 1), (5, 4), (5, 5))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 0], [0, 2], [1, 0], [1, 3], [1, 4], [2, 3], [2, 4], [3, 1], [3, 2], [4, 0], [4, 2], [4, 3], [5, 0], [5, 1], [5, 2], [5, 3], [5, 5]], "position": [2, 5], "goal": [4, 5]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.48, "south": 0.52}, "solvable": {"false": 0.30000000000000004, "true": 0.7}, "value": {"0": 0.09, "1": 0.47000000000000003, "2": 0.31, "3": 0.13}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:b173d428b7997351e0bd9e4f", "state_id": "grid_navigation:b173d428b7997351e0bd9e4f", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. G . # # .\n. . A # # .\n# . # . . .\n. # . . # #\n# . . # . #\n. # . . # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"west": "Move one cell west.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"west": 0.5, "north": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(0, 1):walls=((0, 3), (0, 4), (1, 3), (1, 4), (2, 0), (2, 2), (3, 1), (3, 4), (3, 5), (4, 0), (4, 3), (4, 5), (5, 1), (5, 4))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 3], [0, 4], [1, 3], [1, 4], [2, 0], [2, 2], [3, 1], [3, 4], [3, 5], [4, 0], [4, 3], [4, 5], [5, 1], [5, 4]], "position": [1, 2], "goal": [0, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.72, "west": 0.28}, "solvable": {"false": 0.22999999999999998, "true": 0.77}, "value": {"0": 0.41000000000000003, "1": 0.38, "2": 0.12, "3": 0.09}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:d8eb6b7ec3edcefa1b307696", "state_id": "tic_tac_toe:d8eb6b7ec3edcefa1b307696", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\n. X .\n. . .\n. O X\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_3": "Place O in row 1, column 3 (cell 3).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_5": "Place O in row 2, column 2 (cell 5).", "cell_7": "Place O in row 3, column 1 (cell 7).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_1": "Place O in row 1, column 1 (cell 1)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_1", "solvable": false, "value": 1}, "gold_probs": {"action": {"cell_3": 0.5, "cell_4": 0.0, "cell_5": 0.0, "cell_7": 0.0, "cell_6": 0.0, "cell_1": 0.5}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_1", "cell_3"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:...O.XX..:O", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": ".X.....OX", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_6": 0.14, "cell_5": 0.48, "cell_3": 0.09, "cell_7": 0.12, "cell_4": 0.08, "cell_1": 0.09}, "solvable": {"false": 0.9, "true": 0.1}, "value": {"0": 0.33, "1": 0.51, "2": 0.16}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:40c6304cdec38387b1ea5c74", "state_id": "grid_navigation:40c6304cdec38387b1ea5c74", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. # . # . .\n. . . # A .\n# # # . . .\n. . . . . .\n. . . G . .\n. . . . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "east": "Move one cell east.", "north": "Move one cell north."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"south": 1.0, "east": 0.0, "north": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(1, 2):walls=((3, 3), (3, 4), (3, 5), (4, 2), (5, 2), (5, 4))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 1], [0, 3], [1, 3], [2, 0], [2, 1], [2, 2]], "position": [1, 4], "goal": [4, 3]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.03, "south": 0.9, "east": 0.07}, "solvable": {"false": 0.13, "true": 0.87}, "value": {"0": 0.01, "1": 0.45, "2": 0.51, "3": 0.03}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:36006163a5a5e5290702b5e6", "state_id": "grid_navigation:36006163a5a5e5290702b5e6", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . # # .\nG . . . # #\nA . # . # .\n. . . . . #\n# # . . . .\n. . . # # .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "north": "Move one cell north.", "east": "Move one cell east."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "north", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.0, "north": 1.0, "east": 0.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["north"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(0, 1):walls=((0, 4), (1, 4), (2, 2), (3, 0), (3, 5), (4, 0), (4, 1), (4, 2), (4, 5), (5, 1), (5, 3))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 3], [0, 4], [1, 4], [1, 5], [2, 2], [2, 4], [3, 5], [4, 0], [4, 1], [5, 3], [5, 4]], "position": [2, 0], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"east": 0.08, "north": 0.81, "south": 0.11}, "solvable": {"false": 0.19999999999999996, "true": 0.8}, "value": {"0": 0.09, "1": 0.6, "2": 0.26, "3": 0.05}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "tic_tac_toe:ff4dc69b13048fb9bcb2dbf5", "state_id": "tic_tac_toe:ff4dc69b13048fb9bcb2dbf5", "family_id": "tic_tac_toe_minimax_v1", "split": "ood", "state": "New game position:\nTic-tac-toe. X starts; turns alternate. Three in a row wins. Empty=.; cells 1-9 go row by row.\nO . X\n. . .\nX . .\nTo move: O.", "questions": {"action": {"type": "choice", "instructions": "Choose a legal move maximizing win/draw/loss under perfect play by both sides. Earlier wins are not preferred; all equally best moves count.", "criteria": {"cell_2": "Place O in row 1, column 2 (cell 2).", "cell_6": "Place O in row 2, column 3 (cell 6).", "cell_9": "Place O in row 3, column 3 (cell 9).", "cell_8": "Place O in row 3, column 2 (cell 8).", "cell_4": "Place O in row 2, column 1 (cell 4).", "cell_5": "Place O in row 2, column 2 (cell 5)."}}, "solvable": {"type": "boolean", "instructions": "Can the player to move force a win against perfect opposing play?"}, "value": {"type": "score", "instructions": "What is the best attainable outcome for the player to move, assuming perfect play by both sides?", "criteria": ["The player to move loses under perfect play.", "The game is a draw under perfect play.", "The player to move wins under perfect play."]}}, "gold": {"action": "cell_2", "solvable": false, "value": 0}, "gold_probs": {"action": {"cell_2": 0.16666666666666666, "cell_6": 0.16666666666666666, "cell_9": 0.16666666666666666, "cell_8": 0.16666666666666666, "cell_4": 0.16666666666666666, "cell_5": 0.16666666666666666}, "solvable": {"false": 1.0, "true": 0.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["cell_2", "cell_4", "cell_5", "cell_6", "cell_8", "cell_9"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "ttt:..X...X.O:O", "template_id": "tic_tac_toe:ood:en:v1", "language": "en", "environment_state": {"game": "tic_tac_toe", "board": "O.X...X..", "player": "O"}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"cell_4": 0.14, "cell_2": 0.08, "cell_6": 0.04, "cell_8": 0.08, "cell_5": 0.58, "cell_9": 0.08}, "solvable": {"false": 0.91, "true": 0.09}, "value": {"0": 0.28, "1": 0.59, "2": 0.13}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:d76abc6f53c0607a35312ae0", "state_id": "grid_navigation:d76abc6f53c0607a35312ae0", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. A # . . #\nG . # . # .\n. . . . # .\n. # # . . #\n# . . # # .\n. . . # . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"south": "Move one cell south.", "west": "Move one cell west."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 0}, "gold_probs": {"action": {"south": 0.5, "west": 0.5}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 1.0, "1": 0.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south", "west"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(0, 1):walls=((0, 4), (1, 3), (2, 0), (2, 1), (2, 3), (3, 4), (3, 5), (4, 1), (4, 2), (4, 4), (5, 0), (5, 3))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 2], [0, 5], [1, 2], [1, 4], [2, 4], [3, 1], [3, 2], [3, 5], [4, 0], [4, 3], [4, 4], [5, 3]], "position": [0, 1], "goal": [1, 0]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"south": 0.15, "west": 0.85}, "solvable": {"false": 0.27, "true": 0.73}, "value": {"0": 0.45999999999999996, "1": 0.28, "2": 0.16, "3": 0.1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "grid_navigation:89eea23b858d1eccf4598cd4", "state_id": "grid_navigation:89eea23b858d1eccf4598cd4", "family_id": "grid_navigation_bfs_v1", "split": "ood", "state": "New game position:\nGrid navigation. Each legal orthogonal step costs 1; no diagonals. A=agent, G=goal, #=wall, .=open.\n. . . . # #\n# # . . . .\n# # . . # A\n# G . . . .\n. # # . . #\n# . . . . .", "questions": {"action": {"type": "choice", "instructions": "Choose a legal next step on a shortest route to G. If G is unreachable, all legal moves tie. Walls cannot be crossed.", "criteria": {"north": "Move one cell north.", "south": "Move one cell south."}}, "solvable": {"type": "boolean", "instructions": "Can the agent reach G by legal orthogonal moves?"}, "value": {"type": "score", "instructions": "Classify the minimum number of legal steps from A to G; use the unreachable category if no route exists.", "criteria": ["The goal is reachable in one or two steps.", "The goal is reachable in three to five steps.", "The goal is reachable in six or more steps.", "The goal cannot be reached."]}}, "gold": {"action": "south", "solvable": true, "value": 1}, "gold_probs": {"action": {"north": 0.0, "south": 1.0}, "solvable": {"false": 0.0, "true": 1.0}, "value": {"0": 0.0, "1": 1.0, "2": 0.0, "3": 0.0}}, "gold_probs_kind": {"action": "optimal_action_policy", "solvable": "deterministic_truth", "value": "deterministic_truth"}, "optimal_actions": {"action": ["south"]}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "grid:6:goal=(1, 2):walls=((0, 0), (0, 2), (0, 3), (0, 4), (1, 1), (1, 3), (1, 4), (2, 1), (4, 3), (4, 5), (5, 1), (5, 5))", "template_id": "grid_navigation:ood:en:v1", "language": "en", "environment_state": {"game": "grid_navigation", "size": 6, "walls": [[0, 4], [0, 5], [1, 0], [1, 1], [2, 0], [2, 1], [2, 4], [3, 0], [4, 1], [4, 2], [4, 5], [5, 0]], "position": [2, 5], "goal": [3, 1]}}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"action": {"north": 0.56, "south": 0.44}, "solvable": {"false": 0.18999999999999995, "true": 0.81}, "value": {"0": 0.01, "1": 0.21, "2": 0.67, "3": 0.11}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:account:denied:1", "state_id": "support:account:denied:1", "family_id": "support_decisions_v1", "split": "train", "state": "工单主诉：忘记账户密码。退款记录：退款申请已拒绝，未付款。使用影响：次要功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "根据记录，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"account": "账户密码与身份验证", "technical": "软件功能错误与故障", "privacy": "个人资料删除与隐私请求", "shipping": "包裹运输与投递问题"}}, "completed": {"type": "boolean", "instructions": "根据记录，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "根据记录，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "account", "completed": false, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:account:denied:1", "template_id": "support:train:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"privacy": 0, "account": 1, "shipping": 0, "technical": 0}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:documentation:issued:2", "state_id": "support:documentation:issued:2", "family_id": "support_decisions_v1", "split": "train", "state": "Ticket topic: a request for the user manual. Refund record: the refund was paid and completed. Impact: a core feature is blocked with a workaround.", "questions": {"route": {"type": "choice", "instructions": "Using this record, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"privacy": "Personal data deletion and privacy", "technical": "Software errors and broken features", "documentation": "Finding manuals and usage guides"}}, "completed": {"type": "boolean", "instructions": "Using this record, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Using this record, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "documentation", "completed": true, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:documentation:issued:2", "template_id": "support:train:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"technical": 0, "privacy": 0, "documentation": 1}, "completed": {"false": 0.12, "true": 0.88}, "severity": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:technical:requested:2", "state_id": "support:technical:requested:2", "family_id": "support_decisions_v1", "split": "train", "state": "工单主诉：软件按钮报错。退款记录：已申请退款，尚未处理。使用影响：核心功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "根据记录，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"account": "账户密码与身份验证", "technical": "软件功能错误与故障"}}, "completed": {"type": "boolean", "instructions": "根据记录，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "根据记录，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "technical", "completed": false, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:technical:requested:2", "template_id": "support:train:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"account": 0, "technical": 1}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:privacy:none:2", "state_id": "support:privacy:none:2", "family_id": "support_decisions_v1", "split": "train", "state": "Ticket topic: a request to delete personal data. Refund record: no refund was requested. Impact: a core feature is blocked with a workaround.", "questions": {"route": {"type": "choice", "instructions": "Using this record, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"shipping": "Parcel transport and delivery", "account": "Passwords and identity checks", "privacy": "Personal data deletion and privacy", "technical": "Software errors and broken features", "billing": "Charges, payments, and refunds"}}, "completed": {"type": "boolean", "instructions": "Using this record, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Using this record, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "privacy", "completed": false, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:privacy:none:2", "template_id": "support:train:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"account": 0, "shipping": 0, "technical": 0, "privacy": 1, "billing": 0}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:privacy:issued:3", "state_id": "support:privacy:issued:3", "family_id": "support_decisions_v1", "split": "train", "state": "Ticket topic: a request to delete personal data. Refund record: the refund was paid and completed. Impact: the entire service is down without a workaround.", "questions": {"route": {"type": "choice", "instructions": "Using this record, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"technical": "Software errors and broken features", "account": "Passwords and identity checks", "privacy": "Personal data deletion and privacy"}}, "completed": {"type": "boolean", "instructions": "Using this record, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Using this record, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "privacy", "completed": true, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:privacy:issued:3", "template_id": "support:train:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"privacy": 1, "account": 0, "technical": 0}, "completed": {"false": 0.20999999999999996, "true": 0.79}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:shipping:issued:2", "state_id": "support:shipping:issued:2", "family_id": "support_decisions_v1", "split": "train", "state": "工单主诉：包裹未送达。退款记录：退款已到账，处理完成。使用影响：核心功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "根据记录，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"shipping": "包裹运输与投递问题", "billing": "扣费、付款与退款问题"}}, "completed": {"type": "boolean", "instructions": "根据记录，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "根据记录，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "shipping", "completed": true, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:shipping:issued:2", "template_id": "support:train:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"shipping": 1, "billing": 0}, "completed": {"false": 0.06999999999999995, "true": 0.93}, "severity": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:shipping:processing:0", "state_id": "support:shipping:processing:0", "family_id": "support_decisions_v1", "split": "train", "state": "工单主诉：包裹未送达。退款记录：退款正在处理中，尚未完成。使用影响：功能正常，没有使用障碍。", "questions": {"route": {"type": "choice", "instructions": "根据记录，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"documentation": "查找说明文档与使用指南", "account": "账户密码与身份验证", "shipping": "包裹运输与投递问题", "billing": "扣费、付款与退款问题"}}, "completed": {"type": "boolean", "instructions": "根据记录，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "根据记录，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "shipping", "completed": false, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:shipping:processing:0", "template_id": "support:train:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"billing": 0, "documentation": 0, "shipping": 1, "account": 0}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:documentation:approved:1", "state_id": "support:documentation:approved:1", "family_id": "support_decisions_v1", "split": "train", "state": "Ticket topic: a request for the user manual. Refund record: the refund was approved but not paid. Impact: a minor feature is blocked with a workaround.", "questions": {"route": {"type": "choice", "instructions": "Using this record, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"shipping": "Parcel transport and delivery", "technical": "Software errors and broken features", "billing": "Charges, payments, and refunds", "account": "Passwords and identity checks", "documentation": "Finding manuals and usage guides"}}, "completed": {"type": "boolean", "instructions": "Using this record, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Using this record, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "documentation", "completed": false, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:documentation:approved:1", "template_id": "support:train:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"documentation": 1, "billing": 0, "shipping": 0, "technical": 0, "account": 0}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:technical:none:1", "state_id": "support:technical:none:1", "family_id": "support_decisions_v1", "split": "train", "state": "工单主诉：软件按钮报错。退款记录：尚未提出退款申请。使用影响：次要功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "根据记录，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"account": "账户密码与身份验证", "technical": "软件功能错误与故障"}}, "completed": {"type": "boolean", "instructions": "根据记录，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "根据记录，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "technical", "completed": false, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:technical:none:1", "template_id": "support:train:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"account": 0, "technical": 1}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:documentation:approved:3", "state_id": "support:documentation:approved:3", "family_id": "support_decisions_v1", "split": "train", "state": "Ticket topic: a request for the user manual. Refund record: the refund was approved but not paid. Impact: the entire service is down without a workaround.", "questions": {"route": {"type": "choice", "instructions": "Using this record, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"documentation": "Finding manuals and usage guides", "technical": "Software errors and broken features", "privacy": "Personal data deletion and privacy"}}, "completed": {"type": "boolean", "instructions": "Using this record, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Using this record, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "documentation", "completed": false, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:documentation:approved:3", "template_id": "support:train:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"technical": 0, "privacy": 0, "documentation": 1}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:shipping:approved:0", "state_id": "support:shipping:approved:0", "family_id": "support_decisions_v1", "split": "train", "state": "工单主诉：包裹未送达。退款记录：退款已批准，但尚未到账。使用影响：功能正常，没有使用障碍。", "questions": {"route": {"type": "choice", "instructions": "根据记录，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"billing": "扣费、付款与退款问题", "documentation": "查找说明文档与使用指南", "account": "账户密码与身份验证", "shipping": "包裹运输与投递问题"}}, "completed": {"type": "boolean", "instructions": "根据记录，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "根据记录，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "shipping", "completed": false, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:shipping:approved:0", "template_id": "support:train:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"account": 0, "billing": 0, "documentation": 0, "shipping": 1}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:documentation:none:3", "state_id": "support:documentation:none:3", "family_id": "support_decisions_v1", "split": "train", "state": "Ticket topic: a request for the user manual. Refund record: no refund was requested. Impact: the entire service is down without a workaround.", "questions": {"route": {"type": "choice", "instructions": "Using this record, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"account": "Passwords and identity checks", "documentation": "Finding manuals and usage guides", "billing": "Charges, payments, and refunds", "shipping": "Parcel transport and delivery", "technical": "Software errors and broken features"}}, "completed": {"type": "boolean", "instructions": "Using this record, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Using this record, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "documentation", "completed": false, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:documentation:none:3", "template_id": "support:train:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"shipping": 0, "billing": 0, "account": 0, "technical": 0, "documentation": 1}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:billing:issued:0", "state_id": "support:billing:issued:0", "family_id": "support_decisions_v1", "split": "train", "state": "工单主诉：重复扣款。退款记录：退款已到账，处理完成。使用影响：功能正常，没有使用障碍。", "questions": {"route": {"type": "choice", "instructions": "根据记录，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"billing": "扣费、付款与退款问题", "shipping": "包裹运输与投递问题"}}, "completed": {"type": "boolean", "instructions": "根据记录，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "根据记录，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "billing", "completed": true, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:billing:issued:0", "template_id": "support:train:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"shipping": 0, "billing": 1}, "completed": {"false": 0.050000000000000044, "true": 0.95}, "severity": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:billing:issued:3", "state_id": "support:billing:issued:3", "family_id": "support_decisions_v1", "split": "train", "state": "工单主诉：重复扣款。退款记录：退款已到账，处理完成。使用影响：服务完全中断，无替代办法。", "questions": {"route": {"type": "choice", "instructions": "根据记录，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"shipping": "包裹运输与投递问题", "documentation": "查找说明文档与使用指南", "billing": "扣费、付款与退款问题", "privacy": "个人资料删除与隐私请求"}}, "completed": {"type": "boolean", "instructions": "根据记录，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "根据记录，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "billing", "completed": true, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:billing:issued:3", "template_id": "support:train:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"billing": 1, "shipping": 0, "documentation": 0, "privacy": 0}, "completed": {"false": 0.07999999999999996, "true": 0.92}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:shipping:processing:2", "state_id": "support:shipping:processing:2", "family_id": "support_decisions_v1", "split": "train", "state": "Ticket topic: an undelivered parcel. Refund record: the refund is processing and not complete. Impact: a core feature is blocked with a workaround.", "questions": {"route": {"type": "choice", "instructions": "Using this record, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"shipping": "Parcel transport and delivery", "privacy": "Personal data deletion and privacy", "documentation": "Finding manuals and usage guides", "billing": "Charges, payments, and refunds", "account": "Passwords and identity checks"}}, "completed": {"type": "boolean", "instructions": "Using this record, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Using this record, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "shipping", "completed": false, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:shipping:processing:2", "template_id": "support:train:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"shipping": 1, "privacy": 0, "account": 0, "documentation": 0, "billing": 0}, "completed": {"false": 0.96, "true": 0.04}, "severity": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:privacy:denied:0", "state_id": "support:privacy:denied:0", "family_id": "support_decisions_v1", "split": "train", "state": "Ticket topic: a request to delete personal data. Refund record: the refund was denied and not paid. Impact: all features work without obstruction.", "questions": {"route": {"type": "choice", "instructions": "Using this record, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"privacy": "Personal data deletion and privacy", "account": "Passwords and identity checks", "documentation": "Finding manuals and usage guides"}}, "completed": {"type": "boolean", "instructions": "Using this record, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Using this record, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "privacy", "completed": false, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:privacy:denied:0", "template_id": "support:train:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"documentation": 0, "privacy": 1, "account": 0}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:privacy:issued:0", "state_id": "support:privacy:issued:0", "family_id": "support_decisions_v1", "split": "train", "state": "工单主诉：要求删除个人资料。退款记录：退款已到账，处理完成。使用影响：功能正常，没有使用障碍。", "questions": {"route": {"type": "choice", "instructions": "根据记录，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"privacy": "个人资料删除与隐私请求", "account": "账户密码与身份验证"}}, "completed": {"type": "boolean", "instructions": "根据记录，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "根据记录，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "privacy", "completed": true, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:privacy:issued:0", "template_id": "support:train:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"privacy": 1, "account": 0}, "completed": {"false": 0.050000000000000044, "true": 0.95}, "severity": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:billing:requested:1", "state_id": "support:billing:requested:1", "family_id": "support_decisions_v1", "split": "train", "state": "工单主诉：重复扣款。退款记录：已申请退款，尚未处理。使用影响：次要功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "根据记录，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"documentation": "查找说明文档与使用指南", "privacy": "个人资料删除与隐私请求", "shipping": "包裹运输与投递问题", "billing": "扣费、付款与退款问题"}}, "completed": {"type": "boolean", "instructions": "根据记录，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "根据记录，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "billing", "completed": false, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:billing:requested:1", "template_id": "support:train:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"privacy": 0, "documentation": 0, "shipping": 0, "billing": 1}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:billing:approved:1", "state_id": "support:billing:approved:1", "family_id": "support_decisions_v1", "split": "train", "state": "Ticket topic: a duplicate charge. Refund record: the refund was approved but not paid. Impact: a minor feature is blocked with a workaround.", "questions": {"route": {"type": "choice", "instructions": "Using this record, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"documentation": "Finding manuals and usage guides", "billing": "Charges, payments, and refunds", "shipping": "Parcel transport and delivery"}}, "completed": {"type": "boolean", "instructions": "Using this record, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Using this record, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "billing", "completed": false, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:billing:approved:1", "template_id": "support:train:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"shipping": 0, "documentation": 0, "billing": 1}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:documentation:requested:3", "state_id": "support:documentation:requested:3", "family_id": "support_decisions_v1", "split": "train", "state": "Ticket topic: a request for the user manual. Refund record: a refund was requested but not processed. Impact: the entire service is down without a workaround.", "questions": {"route": {"type": "choice", "instructions": "Using this record, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"technical": "Software errors and broken features", "documentation": "Finding manuals and usage guides", "account": "Passwords and identity checks"}}, "completed": {"type": "boolean", "instructions": "Using this record, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Using this record, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "documentation", "completed": false, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:documentation:requested:3", "template_id": "support:train:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"account": 0, "documentation": 1, "technical": 0}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:shipping:requested:2", "state_id": "support:shipping:requested:2", "family_id": "support_decisions_v1", "split": "train", "state": "工单主诉：包裹未送达。退款记录：已申请退款，尚未处理。使用影响：核心功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "根据记录，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"billing": "扣费、付款与退款问题", "shipping": "包裹运输与投递问题"}}, "completed": {"type": "boolean", "instructions": "根据记录，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "根据记录，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "shipping", "completed": false, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:shipping:requested:2", "template_id": "support:train:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"shipping": 1, "billing": 0}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:privacy:requested:2", "state_id": "support:privacy:requested:2", "family_id": "support_decisions_v1", "split": "train", "state": "Ticket topic: a request to delete personal data. Refund record: a refund was requested but not processed. Impact: a core feature is blocked with a workaround.", "questions": {"route": {"type": "choice", "instructions": "Using this record, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"billing": "Charges, payments, and refunds", "privacy": "Personal data deletion and privacy", "documentation": "Finding manuals and usage guides", "technical": "Software errors and broken features", "account": "Passwords and identity checks"}}, "completed": {"type": "boolean", "instructions": "Using this record, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Using this record, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "privacy", "completed": false, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:privacy:requested:2", "template_id": "support:train:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"billing": 0, "documentation": 0, "privacy": 1, "technical": 0, "account": 0}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:account:approved:2", "state_id": "support:account:approved:2", "family_id": "support_decisions_v1", "split": "train", "state": "工单主诉：忘记账户密码。退款记录：退款已批准，但尚未到账。使用影响：核心功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "根据记录，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"technical": "软件功能错误与故障", "account": "账户密码与身份验证"}}, "completed": {"type": "boolean", "instructions": "根据记录，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "根据记录，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "account", "completed": false, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:account:approved:2", "template_id": "support:train:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"account": 1, "technical": 0}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:documentation:none:0", "state_id": "support:documentation:none:0", "family_id": "support_decisions_v1", "split": "train", "state": "Ticket topic: a request for the user manual. Refund record: no refund was requested. Impact: all features work without obstruction.", "questions": {"route": {"type": "choice", "instructions": "Using this record, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"technical": "Software errors and broken features", "privacy": "Personal data deletion and privacy", "shipping": "Parcel transport and delivery", "documentation": "Finding manuals and usage guides", "account": "Passwords and identity checks"}}, "completed": {"type": "boolean", "instructions": "Using this record, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Using this record, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "documentation", "completed": false, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:documentation:none:0", "template_id": "support:train:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"account": 0, "shipping": 0, "technical": 0, "documentation": 1, "privacy": 0}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:billing:processing:1", "state_id": "support:billing:processing:1", "family_id": "support_decisions_v1", "split": "train", "state": "工单主诉：重复扣款。退款记录：退款正在处理中，尚未完成。使用影响：次要功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "根据记录，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"documentation": "查找说明文档与使用指南", "shipping": "包裹运输与投递问题", "billing": "扣费、付款与退款问题", "technical": "软件功能错误与故障"}}, "completed": {"type": "boolean", "instructions": "根据记录，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "根据记录，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "billing", "completed": false, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:billing:processing:1", "template_id": "support:train:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"shipping": 0, "technical": 0, "documentation": 0, "billing": 1}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:technical:denied:1", "state_id": "support:technical:denied:1", "family_id": "support_decisions_v1", "split": "train", "state": "工单主诉：软件按钮报错。退款记录：退款申请已拒绝，未付款。使用影响：次要功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "根据记录，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"billing": "扣费、付款与退款问题", "account": "账户密码与身份验证", "technical": "软件功能错误与故障", "documentation": "查找说明文档与使用指南"}}, "completed": {"type": "boolean", "instructions": "根据记录，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "根据记录，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "technical", "completed": false, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:technical:denied:1", "template_id": "support:train:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"account": 0, "technical": 1, "billing": 0, "documentation": 0}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:shipping:none:3", "state_id": "support:shipping:none:3", "family_id": "support_decisions_v1", "split": "train", "state": "Ticket topic: an undelivered parcel. Refund record: no refund was requested. Impact: the entire service is down without a workaround.", "questions": {"route": {"type": "choice", "instructions": "Using this record, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"shipping": "Parcel transport and delivery", "billing": "Charges, payments, and refunds", "technical": "Software errors and broken features"}}, "completed": {"type": "boolean", "instructions": "Using this record, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Using this record, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "shipping", "completed": false, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:shipping:none:3", "template_id": "support:train:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"technical": 0, "billing": 0, "shipping": 1}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:shipping:issued:0", "state_id": "support:shipping:issued:0", "family_id": "support_decisions_v1", "split": "train", "state": "Ticket topic: an undelivered parcel. Refund record: the refund was paid and completed. Impact: all features work without obstruction.", "questions": {"route": {"type": "choice", "instructions": "Using this record, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"documentation": "Finding manuals and usage guides", "privacy": "Personal data deletion and privacy", "billing": "Charges, payments, and refunds", "shipping": "Parcel transport and delivery", "technical": "Software errors and broken features"}}, "completed": {"type": "boolean", "instructions": "Using this record, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Using this record, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "shipping", "completed": true, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:shipping:issued:0", "template_id": "support:train:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"shipping": 1, "documentation": 0, "privacy": 0, "technical": 0, "billing": 0}, "completed": {"false": 0.12, "true": 0.88}, "severity": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:shipping:denied:0", "state_id": "support:shipping:denied:0", "family_id": "support_decisions_v1", "split": "train", "state": "工单主诉：包裹未送达。退款记录：退款申请已拒绝，未付款。使用影响：功能正常，没有使用障碍。", "questions": {"route": {"type": "choice", "instructions": "根据记录，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"billing": "扣费、付款与退款问题", "shipping": "包裹运输与投递问题"}}, "completed": {"type": "boolean", "instructions": "根据记录，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "根据记录，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "shipping", "completed": false, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:shipping:denied:0", "template_id": "support:train:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"billing": 0, "shipping": 1}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:shipping:requested:3", "state_id": "support:shipping:requested:3", "family_id": "support_decisions_v1", "split": "train", "state": "Ticket topic: an undelivered parcel. Refund record: a refund was requested but not processed. Impact: the entire service is down without a workaround.", "questions": {"route": {"type": "choice", "instructions": "Using this record, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"shipping": "Parcel transport and delivery", "account": "Passwords and identity checks", "billing": "Charges, payments, and refunds"}}, "completed": {"type": "boolean", "instructions": "Using this record, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Using this record, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "shipping", "completed": false, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:shipping:requested:3", "template_id": "support:train:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"billing": 0, "shipping": 1, "account": 0}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:shipping:approved:2", "state_id": "support:shipping:approved:2", "family_id": "support_decisions_v1", "split": "train", "state": "工单主诉：包裹未送达。退款记录：退款已批准，但尚未到账。使用影响：核心功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "根据记录，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"documentation": "查找说明文档与使用指南", "shipping": "包裹运输与投递问题", "billing": "扣费、付款与退款问题", "technical": "软件功能错误与故障"}}, "completed": {"type": "boolean", "instructions": "根据记录，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "根据记录，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "shipping", "completed": false, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:shipping:approved:2", "template_id": "support:train:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"shipping": 1, "technical": 0, "documentation": 0, "billing": 0}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:documentation:issued:1", "state_id": "support:documentation:issued:1", "family_id": "support_decisions_v1", "split": "train", "state": "工单主诉：询问操作手册位置。退款记录：退款已到账，处理完成。使用影响：次要功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "根据记录，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"documentation": "查找说明文档与使用指南", "technical": "软件功能错误与故障"}}, "completed": {"type": "boolean", "instructions": "根据记录，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "根据记录，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "documentation", "completed": true, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:documentation:issued:1", "template_id": "support:train:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"documentation": 1, "technical": 0}, "completed": {"false": 0.050000000000000044, "true": 0.95}, "severity": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:billing:processing:3", "state_id": "support:billing:processing:3", "family_id": "support_decisions_v1", "split": "train", "state": "Ticket topic: a duplicate charge. Refund record: the refund is processing and not complete. Impact: the entire service is down without a workaround.", "questions": {"route": {"type": "choice", "instructions": "Using this record, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"technical": "Software errors and broken features", "shipping": "Parcel transport and delivery", "documentation": "Finding manuals and usage guides", "billing": "Charges, payments, and refunds", "account": "Passwords and identity checks"}}, "completed": {"type": "boolean", "instructions": "Using this record, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Using this record, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "billing", "completed": false, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:billing:processing:3", "template_id": "support:train:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"technical": 0, "documentation": 0, "shipping": 0, "account": 0, "billing": 1}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:technical:denied:2", "state_id": "support:technical:denied:2", "family_id": "support_decisions_v1", "split": "train", "state": "Ticket topic: a software button error. Refund record: the refund was denied and not paid. Impact: a core feature is blocked with a workaround.", "questions": {"route": {"type": "choice", "instructions": "Using this record, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"technical": "Software errors and broken features", "documentation": "Finding manuals and usage guides", "account": "Passwords and identity checks"}}, "completed": {"type": "boolean", "instructions": "Using this record, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Using this record, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "technical", "completed": false, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:technical:denied:2", "template_id": "support:train:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"account": 0, "technical": 1, "documentation": 0}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:privacy:none:0", "state_id": "support:privacy:none:0", "family_id": "support_decisions_v1", "split": "train", "state": "工单主诉：要求删除个人资料。退款记录：尚未提出退款申请。使用影响：功能正常，没有使用障碍。", "questions": {"route": {"type": "choice", "instructions": "根据记录，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"account": "账户密码与身份验证", "billing": "扣费、付款与退款问题", "technical": "软件功能错误与故障", "privacy": "个人资料删除与隐私请求"}}, "completed": {"type": "boolean", "instructions": "根据记录，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "根据记录，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "privacy", "completed": false, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:privacy:none:0", "template_id": "support:train:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"technical": 0, "privacy": 1, "billing": 0, "account": 0}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:shipping:none:0", "state_id": "support:shipping:none:0", "family_id": "support_decisions_v1", "split": "train", "state": "Ticket topic: an undelivered parcel. Refund record: no refund was requested. Impact: all features work without obstruction.", "questions": {"route": {"type": "choice", "instructions": "Using this record, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"billing": "Charges, payments, and refunds", "privacy": "Personal data deletion and privacy", "technical": "Software errors and broken features", "documentation": "Finding manuals and usage guides", "shipping": "Parcel transport and delivery"}}, "completed": {"type": "boolean", "instructions": "Using this record, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Using this record, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "shipping", "completed": false, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:shipping:none:0", "template_id": "support:train:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"billing": 0, "shipping": 1, "technical": 0, "documentation": 0, "privacy": 0}, "completed": {"false": 0.96, "true": 0.04}, "severity": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:account:processing:1", "state_id": "support:account:processing:1", "family_id": "support_decisions_v1", "split": "train", "state": "工单主诉：忘记账户密码。退款记录：退款正在处理中，尚未完成。使用影响：次要功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "根据记录，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"technical": "软件功能错误与故障", "account": "账户密码与身份验证"}}, "completed": {"type": "boolean", "instructions": "根据记录，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "根据记录，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "account", "completed": false, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:account:processing:1", "template_id": "support:train:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"account": 1, "technical": 0}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:technical:processing:1", "state_id": "support:technical:processing:1", "family_id": "support_decisions_v1", "split": "train", "state": "工单主诉：软件按钮报错。退款记录：退款正在处理中，尚未完成。使用影响：次要功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "根据记录，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"documentation": "查找说明文档与使用指南", "account": "账户密码与身份验证", "technical": "软件功能错误与故障", "privacy": "个人资料删除与隐私请求"}}, "completed": {"type": "boolean", "instructions": "根据记录，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "根据记录，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "technical", "completed": false, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:technical:processing:1", "template_id": "support:train:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"privacy": 0, "account": 0, "technical": 1, "documentation": 0}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:technical:processing:3", "state_id": "support:technical:processing:3", "family_id": "support_decisions_v1", "split": "train", "state": "Ticket topic: a software button error. Refund record: the refund is processing and not complete. Impact: the entire service is down without a workaround.", "questions": {"route": {"type": "choice", "instructions": "Using this record, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"account": "Passwords and identity checks", "technical": "Software errors and broken features", "shipping": "Parcel transport and delivery"}}, "completed": {"type": "boolean", "instructions": "Using this record, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Using this record, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "technical", "completed": false, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:technical:processing:3", "template_id": "support:train:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"account": 0, "technical": 1, "shipping": 0}, "completed": {"false": 0.96, "true": 0.04}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:billing:requested:0", "state_id": "support:billing:requested:0", "family_id": "support_decisions_v1", "split": "train", "state": "工单主诉：重复扣款。退款记录：已申请退款，尚未处理。使用影响：功能正常，没有使用障碍。", "questions": {"route": {"type": "choice", "instructions": "根据记录，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"billing": "扣费、付款与退款问题", "shipping": "包裹运输与投递问题"}}, "completed": {"type": "boolean", "instructions": "根据记录，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "根据记录，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "billing", "completed": false, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:billing:requested:0", "template_id": "support:train:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"billing": 1, "shipping": 0}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:documentation:approved:2", "state_id": "support:documentation:approved:2", "family_id": "support_decisions_v1", "split": "train", "state": "Ticket topic: a request for the user manual. Refund record: the refund was approved but not paid. Impact: a core feature is blocked with a workaround.", "questions": {"route": {"type": "choice", "instructions": "Using this record, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"billing": "Charges, payments, and refunds", "documentation": "Finding manuals and usage guides", "technical": "Software errors and broken features"}}, "completed": {"type": "boolean", "instructions": "Using this record, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Using this record, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "documentation", "completed": false, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:documentation:approved:2", "template_id": "support:train:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"billing": 0, "technical": 0, "documentation": 1}, "completed": {"false": 0.96, "true": 0.04}, "severity": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:privacy:issued:1", "state_id": "support:privacy:issued:1", "family_id": "support_decisions_v1", "split": "train", "state": "Ticket topic: a request to delete personal data. Refund record: the refund was paid and completed. Impact: a minor feature is blocked with a workaround.", "questions": {"route": {"type": "choice", "instructions": "Using this record, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"shipping": "Parcel transport and delivery", "technical": "Software errors and broken features", "account": "Passwords and identity checks", "billing": "Charges, payments, and refunds", "privacy": "Personal data deletion and privacy"}}, "completed": {"type": "boolean", "instructions": "Using this record, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Using this record, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "privacy", "completed": true, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:privacy:issued:1", "template_id": "support:train:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"privacy": 1, "shipping": 0, "billing": 0, "account": 0, "technical": 0}, "completed": {"false": 0.08999999999999997, "true": 0.91}, "severity": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:documentation:issued:3", "state_id": "support:documentation:issued:3", "family_id": "support_decisions_v1", "split": "train", "state": "工单主诉：询问操作手册位置。退款记录：退款已到账，处理完成。使用影响：服务完全中断，无替代办法。", "questions": {"route": {"type": "choice", "instructions": "根据记录，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"technical": "软件功能错误与故障", "documentation": "查找说明文档与使用指南", "shipping": "包裹运输与投递问题", "privacy": "个人资料删除与隐私请求"}}, "completed": {"type": "boolean", "instructions": "根据记录，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "根据记录，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "documentation", "completed": true, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:documentation:issued:3", "template_id": "support:train:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"privacy": 0, "shipping": 0, "documentation": 1, "technical": 0}, "completed": {"false": 0.06999999999999995, "true": 0.93}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:billing:approved:0", "state_id": "support:billing:approved:0", "family_id": "support_decisions_v1", "split": "train", "state": "Ticket topic: a duplicate charge. Refund record: the refund was approved but not paid. Impact: all features work without obstruction.", "questions": {"route": {"type": "choice", "instructions": "Using this record, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"documentation": "Finding manuals and usage guides", "technical": "Software errors and broken features", "privacy": "Personal data deletion and privacy", "billing": "Charges, payments, and refunds", "shipping": "Parcel transport and delivery"}}, "completed": {"type": "boolean", "instructions": "Using this record, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Using this record, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "billing", "completed": false, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:billing:approved:0", "template_id": "support:train:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"privacy": 0, "technical": 0, "documentation": 0, "billing": 1, "shipping": 0}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:technical:requested:1", "state_id": "support:technical:requested:1", "family_id": "support_decisions_v1", "split": "train", "state": "工单主诉：软件按钮报错。退款记录：已申请退款，尚未处理。使用影响：次要功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "根据记录，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"technical": "软件功能错误与故障", "account": "账户密码与身份验证"}}, "completed": {"type": "boolean", "instructions": "根据记录，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "根据记录，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "technical", "completed": false, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:technical:requested:1", "template_id": "support:train:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"technical": 1, "account": 0}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:documentation:requested:0", "state_id": "support:documentation:requested:0", "family_id": "support_decisions_v1", "split": "train", "state": "工单主诉：询问操作手册位置。退款记录：已申请退款，尚未处理。使用影响：功能正常，没有使用障碍。", "questions": {"route": {"type": "choice", "instructions": "根据记录，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"documentation": "查找说明文档与使用指南", "technical": "软件功能错误与故障", "billing": "扣费、付款与退款问题", "account": "账户密码与身份验证"}}, "completed": {"type": "boolean", "instructions": "根据记录，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "根据记录，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "documentation", "completed": false, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:documentation:requested:0", "template_id": "support:train:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"account": 0, "technical": 0, "documentation": 1, "billing": 0}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:shipping:issued:3", "state_id": "support:shipping:issued:3", "family_id": "support_decisions_v1", "split": "train", "state": "Ticket topic: an undelivered parcel. Refund record: the refund was paid and completed. Impact: the entire service is down without a workaround.", "questions": {"route": {"type": "choice", "instructions": "Using this record, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"billing": "Charges, payments, and refunds", "account": "Passwords and identity checks", "shipping": "Parcel transport and delivery"}}, "completed": {"type": "boolean", "instructions": "Using this record, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Using this record, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "shipping", "completed": true, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:shipping:issued:3", "template_id": "support:train:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"shipping": 1, "billing": 0, "account": 0}, "completed": {"false": 0.22999999999999998, "true": 0.77}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:technical:processing:0", "state_id": "support:technical:processing:0", "family_id": "support_decisions_v1", "split": "train", "state": "Ticket topic: a software button error. Refund record: the refund is processing and not complete. Impact: all features work without obstruction.", "questions": {"route": {"type": "choice", "instructions": "Using this record, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"documentation": "Finding manuals and usage guides", "billing": "Charges, payments, and refunds", "account": "Passwords and identity checks", "technical": "Software errors and broken features", "privacy": "Personal data deletion and privacy"}}, "completed": {"type": "boolean", "instructions": "Using this record, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Using this record, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "technical", "completed": false, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:technical:processing:0", "template_id": "support:train:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"privacy": 0, "documentation": 0, "billing": 0, "account": 0, "technical": 1}, "completed": {"false": 0.96, "true": 0.04}, "severity": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:shipping:approved:1", "state_id": "support:shipping:approved:1", "family_id": "support_decisions_v1", "split": "train", "state": "工单主诉：包裹未送达。退款记录：退款已批准，但尚未到账。使用影响：次要功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "根据记录，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"billing": "扣费、付款与退款问题", "shipping": "包裹运输与投递问题"}}, "completed": {"type": "boolean", "instructions": "根据记录，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "根据记录，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "shipping", "completed": false, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:shipping:approved:1", "template_id": "support:train:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"shipping": 1, "billing": 0}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:account:requested:1", "state_id": "support:account:requested:1", "family_id": "support_decisions_v1", "split": "train", "state": "Ticket topic: a forgotten account password. Refund record: a refund was requested but not processed. Impact: a minor feature is blocked with a workaround.", "questions": {"route": {"type": "choice", "instructions": "Using this record, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"technical": "Software errors and broken features", "account": "Passwords and identity checks", "shipping": "Parcel transport and delivery"}}, "completed": {"type": "boolean", "instructions": "Using this record, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Using this record, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "account", "completed": false, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:account:requested:1", "template_id": "support:train:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"account": 1, "shipping": 0, "technical": 0}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:billing:none:1", "state_id": "support:billing:none:1", "family_id": "support_decisions_v1", "split": "train", "state": "工单主诉：重复扣款。退款记录：尚未提出退款申请。使用影响：次要功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "根据记录，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"documentation": "查找说明文档与使用指南", "shipping": "包裹运输与投递问题", "account": "账户密码与身份验证", "billing": "扣费、付款与退款问题"}}, "completed": {"type": "boolean", "instructions": "根据记录，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "根据记录，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "billing", "completed": false, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:billing:none:1", "template_id": "support:train:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"shipping": 0, "documentation": 0, "account": 0, "billing": 1}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:billing:issued:2", "state_id": "support:billing:issued:2", "family_id": "support_decisions_v1", "split": "train", "state": "Ticket topic: a duplicate charge. Refund record: the refund was paid and completed. Impact: a core feature is blocked with a workaround.", "questions": {"route": {"type": "choice", "instructions": "Using this record, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"technical": "Software errors and broken features", "documentation": "Finding manuals and usage guides", "billing": "Charges, payments, and refunds", "privacy": "Personal data deletion and privacy", "shipping": "Parcel transport and delivery"}}, "completed": {"type": "boolean", "instructions": "Using this record, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Using this record, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "billing", "completed": true, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:billing:issued:2", "template_id": "support:train:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"shipping": 0, "technical": 0, "privacy": 0, "billing": 1, "documentation": 0}, "completed": {"false": 0.17000000000000004, "true": 0.83}, "severity": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:technical:none:0", "state_id": "support:technical:none:0", "family_id": "support_decisions_v1", "split": "train", "state": "工单主诉：软件按钮报错。退款记录：尚未提出退款申请。使用影响：功能正常，没有使用障碍。", "questions": {"route": {"type": "choice", "instructions": "根据记录，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"technical": "软件功能错误与故障", "account": "账户密码与身份验证"}}, "completed": {"type": "boolean", "instructions": "根据记录，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "根据记录，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "technical", "completed": false, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:technical:none:0", "template_id": "support:train:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"account": 0, "technical": 1}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:billing:none:3", "state_id": "support:billing:none:3", "family_id": "support_decisions_v1", "split": "train", "state": "Ticket topic: a duplicate charge. Refund record: no refund was requested. Impact: the entire service is down without a workaround.", "questions": {"route": {"type": "choice", "instructions": "Using this record, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"account": "Passwords and identity checks", "shipping": "Parcel transport and delivery", "billing": "Charges, payments, and refunds"}}, "completed": {"type": "boolean", "instructions": "Using this record, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Using this record, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "billing", "completed": false, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:billing:none:3", "template_id": "support:train:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"shipping": 0, "billing": 1, "account": 0}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:privacy:requested:0", "state_id": "support:privacy:requested:0", "family_id": "support_decisions_v1", "split": "train", "state": "工单主诉：要求删除个人资料。退款记录：已申请退款，尚未处理。使用影响：功能正常，没有使用障碍。", "questions": {"route": {"type": "choice", "instructions": "根据记录，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"account": "账户密码与身份验证", "privacy": "个人资料删除与隐私请求"}}, "completed": {"type": "boolean", "instructions": "根据记录，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "根据记录，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "privacy", "completed": false, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:privacy:requested:0", "template_id": "support:train:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"privacy": 1, "account": 0}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:shipping:processing:3", "state_id": "support:shipping:processing:3", "family_id": "support_decisions_v1", "split": "train", "state": "工单主诉：包裹未送达。退款记录：退款正在处理中，尚未完成。使用影响：服务完全中断，无替代办法。", "questions": {"route": {"type": "choice", "instructions": "根据记录，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"account": "账户密码与身份验证", "billing": "扣费、付款与退款问题", "privacy": "个人资料删除与隐私请求", "shipping": "包裹运输与投递问题"}}, "completed": {"type": "boolean", "instructions": "根据记录，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "根据记录，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "shipping", "completed": false, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:shipping:processing:3", "template_id": "support:train:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"account": 0, "billing": 0, "shipping": 1, "privacy": 0}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:technical:approved:2", "state_id": "support:technical:approved:2", "family_id": "support_decisions_v1", "split": "train", "state": "Ticket topic: a software button error. Refund record: the refund was approved but not paid. Impact: a core feature is blocked with a workaround.", "questions": {"route": {"type": "choice", "instructions": "Using this record, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"billing": "Charges, payments, and refunds", "account": "Passwords and identity checks", "documentation": "Finding manuals and usage guides", "privacy": "Personal data deletion and privacy", "technical": "Software errors and broken features"}}, "completed": {"type": "boolean", "instructions": "Using this record, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Using this record, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "technical", "completed": false, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:technical:approved:2", "template_id": "support:train:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"privacy": 0, "account": 0, "technical": 1, "documentation": 0, "billing": 0}, "completed": {"false": 0.96, "true": 0.04}, "severity": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:technical:none:2", "state_id": "support:technical:none:2", "family_id": "support_decisions_v1", "split": "train", "state": "Ticket topic: a software button error. Refund record: no refund was requested. Impact: a core feature is blocked with a workaround.", "questions": {"route": {"type": "choice", "instructions": "Using this record, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"technical": "Software errors and broken features", "account": "Passwords and identity checks", "documentation": "Finding manuals and usage guides"}}, "completed": {"type": "boolean", "instructions": "Using this record, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Using this record, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "technical", "completed": false, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:technical:none:2", "template_id": "support:train:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"technical": 1, "documentation": 0, "account": 0}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:account:approved:1", "state_id": "support:account:approved:1", "family_id": "support_decisions_v1", "split": "train", "state": "工单主诉：忘记账户密码。退款记录：退款已批准，但尚未到账。使用影响：次要功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "根据记录，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"account": "账户密码与身份验证", "documentation": "查找说明文档与使用指南", "privacy": "个人资料删除与隐私请求", "technical": "软件功能错误与故障"}}, "completed": {"type": "boolean", "instructions": "根据记录，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "根据记录，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "account", "completed": false, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:account:approved:1", "template_id": "support:train:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"privacy": 0, "technical": 0, "documentation": 0, "account": 1}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:account:approved:0", "state_id": "support:account:approved:0", "family_id": "support_decisions_v1", "split": "train", "state": "Ticket topic: a forgotten account password. Refund record: the refund was approved but not paid. Impact: all features work without obstruction.", "questions": {"route": {"type": "choice", "instructions": "Using this record, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"shipping": "Parcel transport and delivery", "technical": "Software errors and broken features", "billing": "Charges, payments, and refunds", "account": "Passwords and identity checks", "documentation": "Finding manuals and usage guides"}}, "completed": {"type": "boolean", "instructions": "Using this record, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Using this record, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "account", "completed": false, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:account:approved:0", "template_id": "support:train:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"shipping": 0, "billing": 0, "account": 1, "documentation": 0, "technical": 0}, "completed": {"false": 0.96, "true": 0.04}, "severity": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:privacy:approved:2", "state_id": "support:privacy:approved:2", "family_id": "support_decisions_v1", "split": "train", "state": "工单主诉：要求删除个人资料。退款记录：退款已批准，但尚未到账。使用影响：核心功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "根据记录，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"privacy": "个人资料删除与隐私请求", "shipping": "包裹运输与投递问题", "billing": "扣费、付款与退款问题", "account": "账户密码与身份验证"}}, "completed": {"type": "boolean", "instructions": "根据记录，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "根据记录，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "privacy", "completed": false, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:privacy:approved:2", "template_id": "support:train:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"shipping": 0, "privacy": 1, "account": 0, "billing": 0}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:documentation:none:1", "state_id": "support:documentation:none:1", "family_id": "support_decisions_v1", "split": "train", "state": "Ticket topic: a request for the user manual. Refund record: no refund was requested. Impact: a minor feature is blocked with a workaround.", "questions": {"route": {"type": "choice", "instructions": "Using this record, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"technical": "Software errors and broken features", "documentation": "Finding manuals and usage guides", "shipping": "Parcel transport and delivery"}}, "completed": {"type": "boolean", "instructions": "Using this record, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Using this record, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "documentation", "completed": false, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:documentation:none:1", "template_id": "support:train:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"documentation": 1, "shipping": 0, "technical": 0}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:technical:processing:2", "state_id": "support:technical:processing:2", "family_id": "support_decisions_v1", "split": "train", "state": "工单主诉：软件按钮报错。退款记录：退款正在处理中，尚未完成。使用影响：核心功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "根据记录，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"account": "账户密码与身份验证", "technical": "软件功能错误与故障"}}, "completed": {"type": "boolean", "instructions": "根据记录，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "根据记录，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "technical", "completed": false, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:technical:processing:2", "template_id": "support:train:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"technical": 1, "account": 0}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:billing:issued:1", "state_id": "support:billing:issued:1", "family_id": "support_decisions_v1", "split": "train", "state": "Ticket topic: a duplicate charge. Refund record: the refund was paid and completed. Impact: a minor feature is blocked with a workaround.", "questions": {"route": {"type": "choice", "instructions": "Using this record, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"billing": "Charges, payments, and refunds", "documentation": "Finding manuals and usage guides", "privacy": "Personal data deletion and privacy", "technical": "Software errors and broken features", "shipping": "Parcel transport and delivery"}}, "completed": {"type": "boolean", "instructions": "Using this record, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Using this record, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "billing", "completed": true, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:billing:issued:1", "template_id": "support:train:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"billing": 1, "technical": 0, "shipping": 0, "privacy": 0, "documentation": 0}, "completed": {"false": 0.12, "true": 0.88}, "severity": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:account:denied:0", "state_id": "support:account:denied:0", "family_id": "support_decisions_v1", "split": "dev", "state": "客户反馈忘记账户密码；财务状态为退款申请已拒绝，未付款；当前情况是功能正常，没有使用障碍。", "questions": {"route": {"type": "choice", "instructions": "请核对当前事实后判断：只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"technical": "软件功能错误与故障", "account": "账户密码与身份验证"}}, "completed": {"type": "boolean", "instructions": "请核对当前事实后判断：退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "请核对当前事实后判断：按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "account", "completed": false, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:account:denied:0", "template_id": "support:dev:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"account": 1, "technical": 0}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0.95, "1": 0.02, "2": 0.03, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:privacy:denied:2", "state_id": "support:privacy:denied:2", "family_id": "support_decisions_v1", "split": "dev", "state": "The customer reports a request to delete personal data; finance says the refund was denied and not paid; currently a core feature is blocked with a workaround.", "questions": {"route": {"type": "choice", "instructions": "Check the current facts: Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"account": "Passwords and identity checks", "billing": "Charges, payments, and refunds", "privacy": "Personal data deletion and privacy"}}, "completed": {"type": "boolean", "instructions": "Check the current facts: Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Check the current facts: Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "privacy", "completed": false, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:privacy:denied:2", "template_id": "support:dev:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"privacy": 0.99, "account": 0, "billing": 0.01}, "completed": {"false": 0.94, "true": 0.06}, "severity": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:documentation:processing:1", "state_id": "support:documentation:processing:1", "family_id": "support_decisions_v1", "split": "dev", "state": "客户反馈询问操作手册位置；财务状态为退款正在处理中，尚未完成；当前情况是次要功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "请核对当前事实后判断：只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"technical": "软件功能错误与故障", "documentation": "查找说明文档与使用指南", "account": "账户密码与身份验证", "shipping": "包裹运输与投递问题"}}, "completed": {"type": "boolean", "instructions": "请核对当前事实后判断：退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "请核对当前事实后判断：按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "documentation", "completed": false, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:documentation:processing:1", "template_id": "support:dev:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"technical": 0.06, "shipping": 0, "documentation": 0.94, "account": 0}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:billing:denied:0", "state_id": "support:billing:denied:0", "family_id": "support_decisions_v1", "split": "dev", "state": "客户反馈重复扣款；财务状态为退款申请已拒绝，未付款；当前情况是功能正常，没有使用障碍。", "questions": {"route": {"type": "choice", "instructions": "请核对当前事实后判断：只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"shipping": "包裹运输与投递问题", "billing": "扣费、付款与退款问题"}}, "completed": {"type": "boolean", "instructions": "请核对当前事实后判断：退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "请核对当前事实后判断：按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "billing", "completed": false, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:billing:denied:0", "template_id": "support:dev:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"shipping": 0, "billing": 1}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:documentation:issued:0", "state_id": "support:documentation:issued:0", "family_id": "support_decisions_v1", "split": "dev", "state": "The customer reports a request for the user manual; finance says the refund was paid and completed; currently all features work without obstruction.", "questions": {"route": {"type": "choice", "instructions": "Check the current facts: Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"documentation": "Finding manuals and usage guides", "billing": "Charges, payments, and refunds", "technical": "Software errors and broken features"}}, "completed": {"type": "boolean", "instructions": "Check the current facts: Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Check the current facts: Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "documentation", "completed": true, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:documentation:issued:0", "template_id": "support:dev:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"billing": 0.01, "documentation": 0.99, "technical": 0}, "completed": {"false": 0.31000000000000005, "true": 0.69}, "severity": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:account:none:0", "state_id": "support:account:none:0", "family_id": "support_decisions_v1", "split": "dev", "state": "客户反馈忘记账户密码；财务状态为尚未提出退款申请；当前情况是功能正常，没有使用障碍。", "questions": {"route": {"type": "choice", "instructions": "请核对当前事实后判断：只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"privacy": "个人资料删除与隐私请求", "shipping": "包裹运输与投递问题", "account": "账户密码与身份验证", "technical": "软件功能错误与故障"}}, "completed": {"type": "boolean", "instructions": "请核对当前事实后判断：退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "请核对当前事实后判断：按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "account", "completed": false, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:account:none:0", "template_id": "support:dev:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"technical": 0, "privacy": 0, "account": 1, "shipping": 0}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0.95, "1": 0.02, "2": 0.03, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:privacy:issued:2", "state_id": "support:privacy:issued:2", "family_id": "support_decisions_v1", "split": "dev", "state": "客户反馈要求删除个人资料；财务状态为退款已到账，处理完成；当前情况是核心功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "请核对当前事实后判断：只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"account": "账户密码与身份验证", "privacy": "个人资料删除与隐私请求"}}, "completed": {"type": "boolean", "instructions": "请核对当前事实后判断：退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "请核对当前事实后判断：按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "privacy", "completed": true, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:privacy:issued:2", "template_id": "support:dev:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"privacy": 0.99, "account": 0.01}, "completed": {"false": 0.17000000000000004, "true": 0.83}, "severity": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:privacy:requested:3", "state_id": "support:privacy:requested:3", "family_id": "support_decisions_v1", "split": "dev", "state": "The customer reports a request to delete personal data; finance says a refund was requested but not processed; currently the entire service is down without a workaround.", "questions": {"route": {"type": "choice", "instructions": "Check the current facts: Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"account": "Passwords and identity checks", "privacy": "Personal data deletion and privacy", "technical": "Software errors and broken features", "shipping": "Parcel transport and delivery", "documentation": "Finding manuals and usage guides"}}, "completed": {"type": "boolean", "instructions": "Check the current facts: Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Check the current facts: Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "privacy", "completed": false, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:privacy:requested:3", "template_id": "support:dev:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"privacy": 0.92, "documentation": 0, "shipping": 0, "technical": 0.08, "account": 0}, "completed": {"false": 0.96, "true": 0.04}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:account:processing:0", "state_id": "support:account:processing:0", "family_id": "support_decisions_v1", "split": "dev", "state": "The customer reports a forgotten account password; finance says the refund is processing and not complete; currently all features work without obstruction.", "questions": {"route": {"type": "choice", "instructions": "Check the current facts: Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"billing": "Charges, payments, and refunds", "technical": "Software errors and broken features", "documentation": "Finding manuals and usage guides", "privacy": "Personal data deletion and privacy", "account": "Passwords and identity checks"}}, "completed": {"type": "boolean", "instructions": "Check the current facts: Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Check the current facts: Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "account", "completed": false, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:account:processing:0", "template_id": "support:dev:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"privacy": 0, "billing": 0, "account": 1, "documentation": 0, "technical": 0}, "completed": {"false": 0.92, "true": 0.08}, "severity": {"0": 0.74, "1": 0.19, "2": 0.07, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:account:processing:2", "state_id": "support:account:processing:2", "family_id": "support_decisions_v1", "split": "dev", "state": "The customer reports a forgotten account password; finance says the refund is processing and not complete; currently a core feature is blocked with a workaround.", "questions": {"route": {"type": "choice", "instructions": "Check the current facts: Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"shipping": "Parcel transport and delivery", "technical": "Software errors and broken features", "account": "Passwords and identity checks"}}, "completed": {"type": "boolean", "instructions": "Check the current facts: Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Check the current facts: Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "account", "completed": false, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:account:processing:2", "template_id": "support:dev:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"shipping": 0, "account": 0.92, "technical": 0.08}, "completed": {"false": 0.9299999999999999, "true": 0.07}, "severity": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:privacy:requested:1", "state_id": "support:privacy:requested:1", "family_id": "support_decisions_v1", "split": "dev", "state": "客户反馈要求删除个人资料；财务状态为已申请退款，尚未处理；当前情况是次要功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "请核对当前事实后判断：只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"documentation": "查找说明文档与使用指南", "shipping": "包裹运输与投递问题", "privacy": "个人资料删除与隐私请求", "account": "账户密码与身份验证"}}, "completed": {"type": "boolean", "instructions": "请核对当前事实后判断：退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "请核对当前事实后判断：按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "privacy", "completed": false, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:privacy:requested:1", "template_id": "support:dev:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"shipping": 0, "privacy": 1, "account": 0, "documentation": 0}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:account:issued:2", "state_id": "support:account:issued:2", "family_id": "support_decisions_v1", "split": "dev", "state": "The customer reports a forgotten account password; finance says the refund was paid and completed; currently a core feature is blocked with a workaround.", "questions": {"route": {"type": "choice", "instructions": "Check the current facts: Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"documentation": "Finding manuals and usage guides", "shipping": "Parcel transport and delivery", "billing": "Charges, payments, and refunds", "account": "Passwords and identity checks", "technical": "Software errors and broken features"}}, "completed": {"type": "boolean", "instructions": "Check the current facts: Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Check the current facts: Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "account", "completed": true, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:account:issued:2", "template_id": "support:dev:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"billing": 0, "documentation": 0, "account": 0.98, "shipping": 0, "technical": 0.02}, "completed": {"false": 0.38, "true": 0.62}, "severity": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:technical:denied:3", "state_id": "support:technical:denied:3", "family_id": "support_decisions_v1", "split": "dev", "state": "客户反馈软件按钮报错；财务状态为退款申请已拒绝，未付款；当前情况是服务完全中断，无替代办法。", "questions": {"route": {"type": "choice", "instructions": "请核对当前事实后判断：只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"technical": "软件功能错误与故障", "account": "账户密码与身份验证"}}, "completed": {"type": "boolean", "instructions": "请核对当前事实后判断：退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "请核对当前事实后判断：按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "technical", "completed": false, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:technical:denied:3", "template_id": "support:dev:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"account": 0, "technical": 1}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:account:requested:3", "state_id": "support:account:requested:3", "family_id": "support_decisions_v1", "split": "dev", "state": "The customer reports a forgotten account password; finance says a refund was requested but not processed; currently the entire service is down without a workaround.", "questions": {"route": {"type": "choice", "instructions": "Check the current facts: Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"account": "Passwords and identity checks", "documentation": "Finding manuals and usage guides", "technical": "Software errors and broken features"}}, "completed": {"type": "boolean", "instructions": "Check the current facts: Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Check the current facts: Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "account", "completed": false, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:account:requested:3", "template_id": "support:dev:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"account": 0.93, "technical": 0.07, "documentation": 0}, "completed": {"false": 0.95, "true": 0.05}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:account:none:3", "state_id": "support:account:none:3", "family_id": "support_decisions_v1", "split": "dev", "state": "客户反馈忘记账户密码；财务状态为尚未提出退款申请；当前情况是服务完全中断，无替代办法。", "questions": {"route": {"type": "choice", "instructions": "请核对当前事实后判断：只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"technical": "软件功能错误与故障", "privacy": "个人资料删除与隐私请求", "account": "账户密码与身份验证", "shipping": "包裹运输与投递问题"}}, "completed": {"type": "boolean", "instructions": "请核对当前事实后判断：退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "请核对当前事实后判断：按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "account", "completed": false, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:account:none:3", "template_id": "support:dev:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"privacy": 0, "shipping": 0, "account": 1, "technical": 0}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:billing:approved:2", "state_id": "support:billing:approved:2", "family_id": "support_decisions_v1", "split": "calibration", "state": "核查摘要：主要问题是重复扣款。另查得退款已批准，但尚未到账，且核心功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "仅依据这份核查摘要，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"shipping": "包裹运输与投递问题", "billing": "扣费、付款与退款问题"}}, "completed": {"type": "boolean", "instructions": "仅依据这份核查摘要，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "仅依据这份核查摘要，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "billing", "completed": false, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:billing:approved:2", "template_id": "support:calibration:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"shipping": 0, "billing": 1}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:documentation:denied:3", "state_id": "support:documentation:denied:3", "family_id": "support_decisions_v1", "split": "dev", "state": "The customer reports a request for the user manual; finance says the refund was denied and not paid; currently the entire service is down without a workaround.", "questions": {"route": {"type": "choice", "instructions": "Check the current facts: Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"documentation": "Finding manuals and usage guides", "shipping": "Parcel transport and delivery", "account": "Passwords and identity checks", "privacy": "Personal data deletion and privacy", "technical": "Software errors and broken features"}}, "completed": {"type": "boolean", "instructions": "Check the current facts: Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Check the current facts: Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "documentation", "completed": false, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:documentation:denied:3", "template_id": "support:dev:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"technical": 0.32, "privacy": 0, "shipping": 0, "account": 0, "documentation": 0.68}, "completed": {"false": 0.95, "true": 0.05}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:account:requested:2", "state_id": "support:account:requested:2", "family_id": "support_decisions_v1", "split": "calibration", "state": "Review summary: the main issue is a forgotten account password; separately, a refund was requested but not processed, and a core feature is blocked with a workaround.", "questions": {"route": {"type": "choice", "instructions": "Based only on this review, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"account": "Passwords and identity checks", "documentation": "Finding manuals and usage guides", "technical": "Software errors and broken features"}}, "completed": {"type": "boolean", "instructions": "Based only on this review, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Based only on this review, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "account", "completed": false, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:account:requested:2", "template_id": "support:calibration:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"account": 1, "technical": 0, "documentation": 0}, "completed": {"false": 0.96, "true": 0.04}, "severity": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:technical:none:3", "state_id": "support:technical:none:3", "family_id": "support_decisions_v1", "split": "calibration", "state": "核查摘要：主要问题是软件按钮报错。另查得尚未提出退款申请，且服务完全中断，无替代办法。", "questions": {"route": {"type": "choice", "instructions": "仅依据这份核查摘要，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"billing": "扣费、付款与退款问题", "technical": "软件功能错误与故障", "documentation": "查找说明文档与使用指南", "account": "账户密码与身份验证"}}, "completed": {"type": "boolean", "instructions": "仅依据这份核查摘要，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "仅依据这份核查摘要，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "technical", "completed": false, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:technical:none:3", "template_id": "support:calibration:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"billing": 0, "technical": 1, "documentation": 0, "account": 0}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:billing:none:0", "state_id": "support:billing:none:0", "family_id": "support_decisions_v1", "split": "calibration", "state": "Review summary: the main issue is a duplicate charge; separately, no refund was requested, and all features work without obstruction.", "questions": {"route": {"type": "choice", "instructions": "Based only on this review, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"billing": "Charges, payments, and refunds", "privacy": "Personal data deletion and privacy", "technical": "Software errors and broken features", "account": "Passwords and identity checks", "shipping": "Parcel transport and delivery"}}, "completed": {"type": "boolean", "instructions": "Based only on this review, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Based only on this review, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "billing", "completed": false, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:billing:none:0", "template_id": "support:calibration:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"privacy": 0, "shipping": 0, "billing": 1, "account": 0, "technical": 0}, "completed": {"false": 0.94, "true": 0.06}, "severity": {"0": 0.99, "1": 0.01, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:privacy:denied:3", "state_id": "support:privacy:denied:3", "family_id": "support_decisions_v1", "split": "calibration", "state": "核查摘要：主要问题是要求删除个人资料。另查得退款申请已拒绝，未付款，且服务完全中断，无替代办法。", "questions": {"route": {"type": "choice", "instructions": "仅依据这份核查摘要，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"privacy": "个人资料删除与隐私请求", "account": "账户密码与身份验证"}}, "completed": {"type": "boolean", "instructions": "仅依据这份核查摘要，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "仅依据这份核查摘要，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "privacy", "completed": false, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:privacy:denied:3", "template_id": "support:calibration:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"account": 0, "privacy": 1}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:documentation:approved:0", "state_id": "support:documentation:approved:0", "family_id": "support_decisions_v1", "split": "calibration", "state": "Review summary: the main issue is a request for the user manual; separately, the refund was approved but not paid, and all features work without obstruction.", "questions": {"route": {"type": "choice", "instructions": "Based only on this review, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"documentation": "Finding manuals and usage guides", "technical": "Software errors and broken features", "privacy": "Personal data deletion and privacy"}}, "completed": {"type": "boolean", "instructions": "Based only on this review, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Based only on this review, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "documentation", "completed": false, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:documentation:approved:0", "template_id": "support:calibration:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"documentation": 1, "privacy": 0, "technical": 0}, "completed": {"false": 0.95, "true": 0.05}, "severity": {"0": 0.95, "1": 0.03, "2": 0.02, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:technical:approved:1", "state_id": "support:technical:approved:1", "family_id": "support_decisions_v1", "split": "calibration", "state": "核查摘要：主要问题是软件按钮报错。另查得退款已批准，但尚未到账，且次要功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "仅依据这份核查摘要，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"account": "账户密码与身份验证", "technical": "软件功能错误与故障", "documentation": "查找说明文档与使用指南", "billing": "扣费、付款与退款问题"}}, "completed": {"type": "boolean", "instructions": "仅依据这份核查摘要，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "仅依据这份核查摘要，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "technical", "completed": false, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:technical:approved:1", "template_id": "support:calibration:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"documentation": 0, "billing": 0, "technical": 1, "account": 0}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 0.52, "2": 0.48, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:shipping:denied:1", "state_id": "support:shipping:denied:1", "family_id": "support_decisions_v1", "split": "calibration", "state": "Review summary: the main issue is an undelivered parcel; separately, the refund was denied and not paid, and a minor feature is blocked with a workaround.", "questions": {"route": {"type": "choice", "instructions": "Based only on this review, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"privacy": "Personal data deletion and privacy", "documentation": "Finding manuals and usage guides", "billing": "Charges, payments, and refunds", "technical": "Software errors and broken features", "shipping": "Parcel transport and delivery"}}, "completed": {"type": "boolean", "instructions": "Based only on this review, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Based only on this review, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "shipping", "completed": false, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:shipping:denied:1", "template_id": "support:calibration:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"privacy": 0, "technical": 0, "billing": 0, "documentation": 0, "shipping": 1}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 0.51, "2": 0.37, "3": 0.12}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:account:issued:1", "state_id": "support:account:issued:1", "family_id": "support_decisions_v1", "split": "calibration", "state": "核查摘要：主要问题是忘记账户密码。另查得退款已到账，处理完成，且次要功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "仅依据这份核查摘要，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"technical": "软件功能错误与故障", "account": "账户密码与身份验证"}}, "completed": {"type": "boolean", "instructions": "仅依据这份核查摘要，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "仅依据这份核查摘要，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "account", "completed": true, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:account:issued:1", "template_id": "support:calibration:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"account": 1, "technical": 0}, "completed": {"false": 0.06999999999999995, "true": 0.93}, "severity": {"0": 0, "1": 0.62, "2": 0.38, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:privacy:approved:3", "state_id": "support:privacy:approved:3", "family_id": "support_decisions_v1", "split": "calibration", "state": "Review summary: the main issue is a request to delete personal data; separately, the refund was approved but not paid, and the entire service is down without a workaround.", "questions": {"route": {"type": "choice", "instructions": "Based only on this review, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"billing": "Charges, payments, and refunds", "account": "Passwords and identity checks", "privacy": "Personal data deletion and privacy"}}, "completed": {"type": "boolean", "instructions": "Based only on this review, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Based only on this review, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "privacy", "completed": false, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:privacy:approved:3", "template_id": "support:calibration:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"privacy": 1, "account": 0, "billing": 0}, "completed": {"false": 0.96, "true": 0.04}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:shipping:none:1", "state_id": "support:shipping:none:1", "family_id": "support_decisions_v1", "split": "calibration", "state": "核查摘要：主要问题是包裹未送达。另查得尚未提出退款申请，且次要功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "仅依据这份核查摘要，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"account": "账户密码与身份验证", "shipping": "包裹运输与投递问题", "privacy": "个人资料删除与隐私请求", "billing": "扣费、付款与退款问题"}}, "completed": {"type": "boolean", "instructions": "仅依据这份核查摘要，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "仅依据这份核查摘要，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "shipping", "completed": false, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:shipping:none:1", "template_id": "support:calibration:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"shipping": 1, "account": 0, "privacy": 0, "billing": 0}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0, "1": 0.56, "2": 0.31, "3": 0.13}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:account:issued:3", "state_id": "support:account:issued:3", "family_id": "support_decisions_v1", "split": "calibration", "state": "Review summary: the main issue is a forgotten account password; separately, the refund was paid and completed, and the entire service is down without a workaround.", "questions": {"route": {"type": "choice", "instructions": "Based only on this review, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"technical": "Software errors and broken features", "billing": "Charges, payments, and refunds", "shipping": "Parcel transport and delivery", "privacy": "Personal data deletion and privacy", "account": "Passwords and identity checks"}}, "completed": {"type": "boolean", "instructions": "Based only on this review, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Based only on this review, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "account", "completed": true, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:account:issued:3", "template_id": "support:calibration:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"privacy": 0, "account": 1, "technical": 0, "shipping": 0, "billing": 0}, "completed": {"false": 0.09999999999999998, "true": 0.9}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:technical:requested:3", "state_id": "support:technical:requested:3", "family_id": "support_decisions_v1", "split": "calibration", "state": "核查摘要：主要问题是软件按钮报错。另查得已申请退款，尚未处理，且服务完全中断，无替代办法。", "questions": {"route": {"type": "choice", "instructions": "仅依据这份核查摘要，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"account": "账户密码与身份验证", "technical": "软件功能错误与故障"}}, "completed": {"type": "boolean", "instructions": "仅依据这份核查摘要，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "仅依据这份核查摘要，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "technical", "completed": false, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:technical:requested:3", "template_id": "support:calibration:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"technical": 1, "account": 0}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:account:denied:3", "state_id": "support:account:denied:3", "family_id": "support_decisions_v1", "split": "calibration", "state": "核查摘要：主要问题是忘记账户密码。另查得退款申请已拒绝，未付款，且服务完全中断，无替代办法。", "questions": {"route": {"type": "choice", "instructions": "仅依据这份核查摘要，只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"billing": "扣费、付款与退款问题", "account": "账户密码与身份验证", "shipping": "包裹运输与投递问题", "technical": "软件功能错误与故障"}}, "completed": {"type": "boolean", "instructions": "仅依据这份核查摘要，退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "仅依据这份核查摘要，按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "account", "completed": false, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:account:denied:3", "template_id": "support:calibration:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"technical": 0, "shipping": 0, "billing": 0, "account": 1}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:billing:none:2", "state_id": "support:billing:none:2", "family_id": "support_decisions_v1", "split": "calibration", "state": "Review summary: the main issue is a duplicate charge; separately, no refund was requested, and a core feature is blocked with a workaround.", "questions": {"route": {"type": "choice", "instructions": "Based only on this review, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"billing": "Charges, payments, and refunds", "privacy": "Personal data deletion and privacy", "shipping": "Parcel transport and delivery"}}, "completed": {"type": "boolean", "instructions": "Based only on this review, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Based only on this review, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "billing", "completed": false, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:billing:none:2", "template_id": "support:calibration:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"shipping": 0, "privacy": 0, "billing": 1}, "completed": {"false": 0.94, "true": 0.06}, "severity": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:documentation:denied:2", "state_id": "support:documentation:denied:2", "family_id": "support_decisions_v1", "split": "calibration", "state": "Review summary: the main issue is a request for the user manual; separately, the refund was denied and not paid, and a core feature is blocked with a workaround.", "questions": {"route": {"type": "choice", "instructions": "Based only on this review, Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"documentation": "Finding manuals and usage guides", "shipping": "Parcel transport and delivery", "account": "Passwords and identity checks", "billing": "Charges, payments, and refunds", "technical": "Software errors and broken features"}}, "completed": {"type": "boolean", "instructions": "Based only on this review, Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Based only on this review, Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "documentation", "completed": false, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:documentation:denied:2", "template_id": "support:calibration:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"technical": 0, "shipping": 0, "billing": 0, "account": 0, "documentation": 1}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:privacy:none:3", "state_id": "support:privacy:none:3", "family_id": "support_decisions_v1", "split": "test", "state": "请看独立记录——问题：要求删除个人资料；退款：尚未提出退款申请；功能：服务完全中断，无替代办法。", "questions": {"route": {"type": "choice", "instructions": "读取上述独立记录并回答：只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"privacy": "个人资料删除与隐私请求", "account": "账户密码与身份验证"}}, "completed": {"type": "boolean", "instructions": "读取上述独立记录并回答：退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "读取上述独立记录并回答：按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "privacy", "completed": false, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:privacy:none:3", "template_id": "support:test:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"account": 0, "privacy": 1}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:documentation:denied:1", "state_id": "support:documentation:denied:1", "family_id": "support_decisions_v1", "split": "test", "state": "Independent log — topic: a request for the user manual; refund: the refund was denied and not paid; functionality: a minor feature is blocked with a workaround.", "questions": {"route": {"type": "choice", "instructions": "Read the independent log and answer: Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"documentation": "Finding manuals and usage guides", "technical": "Software errors and broken features", "shipping": "Parcel transport and delivery"}}, "completed": {"type": "boolean", "instructions": "Read the independent log and answer: Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Read the independent log and answer: Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "documentation", "completed": false, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:documentation:denied:1", "template_id": "support:test:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"shipping": 0, "technical": 0, "documentation": 1}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:billing:denied:2", "state_id": "support:billing:denied:2", "family_id": "support_decisions_v1", "split": "test", "state": "请看独立记录——问题：重复扣款；退款：退款申请已拒绝，未付款；功能：核心功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "读取上述独立记录并回答：只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"privacy": "个人资料删除与隐私请求", "billing": "扣费、付款与退款问题", "shipping": "包裹运输与投递问题", "account": "账户密码与身份验证"}}, "completed": {"type": "boolean", "instructions": "读取上述独立记录并回答：退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "读取上述独立记录并回答：按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "billing", "completed": false, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:billing:denied:2", "template_id": "support:test:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"shipping": 0, "privacy": 0, "billing": 0.98, "account": 0.02}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:shipping:requested:0", "state_id": "support:shipping:requested:0", "family_id": "support_decisions_v1", "split": "test", "state": "Independent log — topic: an undelivered parcel; refund: a refund was requested but not processed; functionality: all features work without obstruction.", "questions": {"route": {"type": "choice", "instructions": "Read the independent log and answer: Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"technical": "Software errors and broken features", "billing": "Charges, payments, and refunds", "shipping": "Parcel transport and delivery", "documentation": "Finding manuals and usage guides", "account": "Passwords and identity checks"}}, "completed": {"type": "boolean", "instructions": "Read the independent log and answer: Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Read the independent log and answer: Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "shipping", "completed": false, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:shipping:requested:0", "template_id": "support:test:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"shipping": 1, "billing": 0, "account": 0, "documentation": 0, "technical": 0}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0.8, "1": 0.07, "2": 0.13, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:privacy:approved:0", "state_id": "support:privacy:approved:0", "family_id": "support_decisions_v1", "split": "test", "state": "请看独立记录——问题：要求删除个人资料；退款：退款已批准，但尚未到账；功能：功能正常，没有使用障碍。", "questions": {"route": {"type": "choice", "instructions": "读取上述独立记录并回答：只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"privacy": "个人资料删除与隐私请求", "account": "账户密码与身份验证"}}, "completed": {"type": "boolean", "instructions": "读取上述独立记录并回答：退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "读取上述独立记录并回答：按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "privacy", "completed": false, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:privacy:approved:0", "template_id": "support:test:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"account": 0, "privacy": 1}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0.95, "1": 0.04, "2": 0.01, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:technical:approved:0", "state_id": "support:technical:approved:0", "family_id": "support_decisions_v1", "split": "test", "state": "Independent log — topic: a software button error; refund: the refund was approved but not paid; functionality: all features work without obstruction.", "questions": {"route": {"type": "choice", "instructions": "Read the independent log and answer: Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"technical": "Software errors and broken features", "account": "Passwords and identity checks", "shipping": "Parcel transport and delivery"}}, "completed": {"type": "boolean", "instructions": "Read the independent log and answer: Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Read the independent log and answer: Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "technical", "completed": false, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:technical:approved:0", "template_id": "support:test:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"shipping": 0, "account": 0, "technical": 1}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0.88, "1": 0.1, "2": 0.02, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:billing:processing:2", "state_id": "support:billing:processing:2", "family_id": "support_decisions_v1", "split": "test", "state": "Independent log — topic: a duplicate charge; refund: the refund is processing and not complete; functionality: a core feature is blocked with a workaround.", "questions": {"route": {"type": "choice", "instructions": "Read the independent log and answer: Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"billing": "Charges, payments, and refunds", "privacy": "Personal data deletion and privacy", "documentation": "Finding manuals and usage guides", "shipping": "Parcel transport and delivery", "technical": "Software errors and broken features"}}, "completed": {"type": "boolean", "instructions": "Read the independent log and answer: Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Read the independent log and answer: Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "billing", "completed": false, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:billing:processing:2", "template_id": "support:test:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"billing": 0.99, "shipping": 0, "technical": 0.01, "documentation": 0, "privacy": 0}, "completed": {"false": 0.96, "true": 0.04}, "severity": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:privacy:processing:3", "state_id": "support:privacy:processing:3", "family_id": "support_decisions_v1", "split": "test", "state": "请看独立记录——问题：要求删除个人资料；退款：退款正在处理中，尚未完成；功能：服务完全中断，无替代办法。", "questions": {"route": {"type": "choice", "instructions": "读取上述独立记录并回答：只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"documentation": "查找说明文档与使用指南", "shipping": "包裹运输与投递问题", "privacy": "个人资料删除与隐私请求", "account": "账户密码与身份验证"}}, "completed": {"type": "boolean", "instructions": "读取上述独立记录并回答：退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "读取上述独立记录并回答：按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "privacy", "completed": false, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:privacy:processing:3", "template_id": "support:test:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"shipping": 0, "privacy": 0.98, "account": 0.02, "documentation": 0}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:technical:requested:0", "state_id": "support:technical:requested:0", "family_id": "support_decisions_v1", "split": "test", "state": "请看独立记录——问题：软件按钮报错；退款：已申请退款，尚未处理；功能：功能正常，没有使用障碍。", "questions": {"route": {"type": "choice", "instructions": "读取上述独立记录并回答：只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"account": "账户密码与身份验证", "technical": "软件功能错误与故障"}}, "completed": {"type": "boolean", "instructions": "读取上述独立记录并回答：退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "读取上述独立记录并回答：按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "technical", "completed": false, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:technical:requested:0", "template_id": "support:test:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"technical": 1, "account": 0}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0.8099999999999999, "1": 0.18, "2": 0.01, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:account:approved:3", "state_id": "support:account:approved:3", "family_id": "support_decisions_v1", "split": "test", "state": "Independent log — topic: a forgotten account password; refund: the refund was approved but not paid; functionality: the entire service is down without a workaround.", "questions": {"route": {"type": "choice", "instructions": "Read the independent log and answer: Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"account": "Passwords and identity checks", "technical": "Software errors and broken features", "documentation": "Finding manuals and usage guides"}}, "completed": {"type": "boolean", "instructions": "Read the independent log and answer: Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Read the independent log and answer: Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "account", "completed": false, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:account:approved:3", "template_id": "support:test:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"account": 1, "documentation": 0, "technical": 0}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:billing:processing:0", "state_id": "support:billing:processing:0", "family_id": "support_decisions_v1", "split": "test", "state": "请看独立记录——问题：重复扣款；退款：退款正在处理中，尚未完成；功能：功能正常，没有使用障碍。", "questions": {"route": {"type": "choice", "instructions": "读取上述独立记录并回答：只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"billing": "扣费、付款与退款问题", "account": "账户密码与身份验证", "privacy": "个人资料删除与隐私请求", "shipping": "包裹运输与投递问题"}}, "completed": {"type": "boolean", "instructions": "读取上述独立记录并回答：退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "读取上述独立记录并回答：按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "billing", "completed": false, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:billing:processing:0", "template_id": "support:test:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"privacy": 0, "account": 0, "billing": 1, "shipping": 0}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0.92, "1": 0.05, "2": 0.03, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:privacy:processing:1", "state_id": "support:privacy:processing:1", "family_id": "support_decisions_v1", "split": "test", "state": "Independent log — topic: a request to delete personal data; refund: the refund is processing and not complete; functionality: a minor feature is blocked with a workaround.", "questions": {"route": {"type": "choice", "instructions": "Read the independent log and answer: Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"billing": "Charges, payments, and refunds", "privacy": "Personal data deletion and privacy", "shipping": "Parcel transport and delivery", "technical": "Software errors and broken features", "account": "Passwords and identity checks"}}, "completed": {"type": "boolean", "instructions": "Read the independent log and answer: Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Read the independent log and answer: Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "privacy", "completed": false, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:privacy:processing:1", "template_id": "support:test:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"account": 0, "shipping": 0, "technical": 0, "privacy": 1, "billing": 0}, "completed": {"false": 0.96, "true": 0.04}, "severity": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:shipping:denied:3", "state_id": "support:shipping:denied:3", "family_id": "support_decisions_v1", "split": "test", "state": "请看独立记录——问题：包裹未送达；退款：退款申请已拒绝，未付款；功能：服务完全中断，无替代办法。", "questions": {"route": {"type": "choice", "instructions": "读取上述独立记录并回答：只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"billing": "扣费、付款与退款问题", "shipping": "包裹运输与投递问题"}}, "completed": {"type": "boolean", "instructions": "读取上述独立记录并回答：退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "读取上述独立记录并回答：按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "shipping", "completed": false, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:shipping:denied:3", "template_id": "support:test:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"shipping": 1, "billing": 0}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:technical:issued:1", "state_id": "support:technical:issued:1", "family_id": "support_decisions_v1", "split": "test", "state": "请看独立记录——问题：软件按钮报错；退款：退款已到账，处理完成；功能：次要功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "读取上述独立记录并回答：只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"technical": "软件功能错误与故障", "privacy": "个人资料删除与隐私请求", "billing": "扣费、付款与退款问题", "account": "账户密码与身份验证"}}, "completed": {"type": "boolean", "instructions": "读取上述独立记录并回答：退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "读取上述独立记录并回答：按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "technical", "completed": true, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:technical:issued:1", "template_id": "support:test:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"privacy": 0, "account": 0, "billing": 0.01, "technical": 0.99}, "completed": {"false": 0.06000000000000005, "true": 0.94}, "severity": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:billing:approved:3", "state_id": "support:billing:approved:3", "family_id": "support_decisions_v1", "split": "test", "state": "Independent log — topic: a duplicate charge; refund: the refund was approved but not paid; functionality: the entire service is down without a workaround.", "questions": {"route": {"type": "choice", "instructions": "Read the independent log and answer: Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"shipping": "Parcel transport and delivery", "billing": "Charges, payments, and refunds", "technical": "Software errors and broken features"}}, "completed": {"type": "boolean", "instructions": "Read the independent log and answer: Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Read the independent log and answer: Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "billing", "completed": false, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:billing:approved:3", "template_id": "support:test:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"shipping": 0, "billing": 1, "technical": 0}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:shipping:approved:3", "state_id": "support:shipping:approved:3", "family_id": "support_decisions_v1", "split": "test", "state": "Independent log — topic: an undelivered parcel; refund: the refund was approved but not paid; functionality: the entire service is down without a workaround.", "questions": {"route": {"type": "choice", "instructions": "Read the independent log and answer: Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"documentation": "Finding manuals and usage guides", "billing": "Charges, payments, and refunds", "shipping": "Parcel transport and delivery"}}, "completed": {"type": "boolean", "instructions": "Read the independent log and answer: Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Read the independent log and answer: Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "shipping", "completed": false, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:shipping:approved:3", "template_id": "support:test:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"billing": 0, "shipping": 1, "documentation": 0}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:account:denied:2", "state_id": "support:account:denied:2", "family_id": "support_decisions_v1", "split": "test", "state": "Independent log — topic: a forgotten account password; refund: the refund was denied and not paid; functionality: a core feature is blocked with a workaround.", "questions": {"route": {"type": "choice", "instructions": "Read the independent log and answer: Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"documentation": "Finding manuals and usage guides", "privacy": "Personal data deletion and privacy", "technical": "Software errors and broken features", "account": "Passwords and identity checks", "billing": "Charges, payments, and refunds"}}, "completed": {"type": "boolean", "instructions": "Read the independent log and answer: Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Read the independent log and answer: Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "account", "completed": false, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:account:denied:2", "template_id": "support:test:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"technical": 0, "privacy": 0, "account": 1, "documentation": 0, "billing": 0}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:account:none:2", "state_id": "support:account:none:2", "family_id": "support_decisions_v1", "split": "test", "state": "请看独立记录——问题：忘记账户密码；退款：尚未提出退款申请；功能：核心功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "读取上述独立记录并回答：只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"account": "账户密码与身份验证", "technical": "软件功能错误与故障"}}, "completed": {"type": "boolean", "instructions": "读取上述独立记录并回答：退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "读取上述独立记录并回答：按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "account", "completed": false, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:account:none:2", "template_id": "support:test:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"account": 1, "technical": 0}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:privacy:processing:2", "state_id": "support:privacy:processing:2", "family_id": "support_decisions_v1", "split": "test", "state": "请看独立记录——问题：要求删除个人资料；退款：退款正在处理中，尚未完成；功能：核心功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "读取上述独立记录并回答：只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"technical": "软件功能错误与故障", "documentation": "查找说明文档与使用指南", "privacy": "个人资料删除与隐私请求", "account": "账户密码与身份验证"}}, "completed": {"type": "boolean", "instructions": "读取上述独立记录并回答：退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "读取上述独立记录并回答：按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "privacy", "completed": false, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:privacy:processing:2", "template_id": "support:test:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"account": 0.01, "technical": 0.04, "documentation": 0, "privacy": 0.95}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:privacy:denied:1", "state_id": "support:privacy:denied:1", "family_id": "support_decisions_v1", "split": "test", "state": "Independent log — topic: a request to delete personal data; refund: the refund was denied and not paid; functionality: a minor feature is blocked with a workaround.", "questions": {"route": {"type": "choice", "instructions": "Read the independent log and answer: Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"privacy": "Personal data deletion and privacy", "account": "Passwords and identity checks", "shipping": "Parcel transport and delivery", "documentation": "Finding manuals and usage guides", "technical": "Software errors and broken features"}}, "completed": {"type": "boolean", "instructions": "Read the independent log and answer: Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Read the independent log and answer: Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "privacy", "completed": false, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:privacy:denied:1", "template_id": "support:test:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"technical": 0, "shipping": 0, "documentation": 0, "account": 0, "privacy": 1}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 0.99, "2": 0.01, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:technical:issued:0", "state_id": "support:technical:issued:0", "family_id": "support_decisions_v1", "split": "test", "state": "请看独立记录——问题：软件按钮报错；退款：退款已到账，处理完成；功能：功能正常，没有使用障碍。", "questions": {"route": {"type": "choice", "instructions": "读取上述独立记录并回答：只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"account": "账户密码与身份验证", "technical": "软件功能错误与故障"}}, "completed": {"type": "boolean", "instructions": "读取上述独立记录并回答：退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "读取上述独立记录并回答：按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "technical", "completed": true, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:technical:issued:0", "template_id": "support:test:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"account": 0, "technical": 1}, "completed": {"false": 0.06000000000000005, "true": 0.94}, "severity": {"0": 0.8, "1": 0.19, "2": 0.01, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:privacy:none:1", "state_id": "support:privacy:none:1", "family_id": "support_decisions_v1", "split": "test", "state": "请看独立记录——问题：要求删除个人资料；退款：尚未提出退款申请；功能：次要功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "读取上述独立记录并回答：只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"billing": "扣费、付款与退款问题", "privacy": "个人资料删除与隐私请求", "account": "账户密码与身份验证", "documentation": "查找说明文档与使用指南"}}, "completed": {"type": "boolean", "instructions": "读取上述独立记录并回答：退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "读取上述独立记录并回答：按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "privacy", "completed": false, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:privacy:none:1", "template_id": "support:test:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"account": 0, "privacy": 1, "documentation": 0, "billing": 0}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:shipping:denied:2", "state_id": "support:shipping:denied:2", "family_id": "support_decisions_v1", "split": "test", "state": "Independent log — topic: an undelivered parcel; refund: the refund was denied and not paid; functionality: a core feature is blocked with a workaround.", "questions": {"route": {"type": "choice", "instructions": "Read the independent log and answer: Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"shipping": "Parcel transport and delivery", "billing": "Charges, payments, and refunds", "privacy": "Personal data deletion and privacy"}}, "completed": {"type": "boolean", "instructions": "Read the independent log and answer: Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Read the independent log and answer: Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "shipping", "completed": false, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:shipping:denied:2", "template_id": "support:test:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"billing": 0, "privacy": 0, "shipping": 1}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:billing:denied:3", "state_id": "support:billing:denied:3", "family_id": "support_decisions_v1", "split": "test", "state": "Independent log — topic: a duplicate charge; refund: the refund was denied and not paid; functionality: the entire service is down without a workaround.", "questions": {"route": {"type": "choice", "instructions": "Read the independent log and answer: Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"billing": "Charges, payments, and refunds", "privacy": "Personal data deletion and privacy", "documentation": "Finding manuals and usage guides", "technical": "Software errors and broken features", "shipping": "Parcel transport and delivery"}}, "completed": {"type": "boolean", "instructions": "Read the independent log and answer: Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Read the independent log and answer: Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "billing", "completed": false, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:billing:denied:3", "template_id": "support:test:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"billing": 0.99, "shipping": 0, "documentation": 0, "privacy": 0, "technical": 0.01}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:documentation:denied:0", "state_id": "support:documentation:denied:0", "family_id": "support_decisions_v1", "split": "test", "state": "Independent log — topic: a request for the user manual; refund: the refund was denied and not paid; functionality: all features work without obstruction.", "questions": {"route": {"type": "choice", "instructions": "Read the independent log and answer: Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"account": "Passwords and identity checks", "documentation": "Finding manuals and usage guides", "technical": "Software errors and broken features"}}, "completed": {"type": "boolean", "instructions": "Read the independent log and answer: Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Read the independent log and answer: Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "documentation", "completed": false, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:documentation:denied:0", "template_id": "support:test:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"technical": 0, "account": 0, "documentation": 1}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0.98, "1": 0.01, "2": 0.01, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:documentation:requested:1", "state_id": "support:documentation:requested:1", "family_id": "support_decisions_v1", "split": "test", "state": "请看独立记录——问题：询问操作手册位置；退款：已申请退款，尚未处理；功能：次要功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "读取上述独立记录并回答：只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"billing": "扣费、付款与退款问题", "documentation": "查找说明文档与使用指南", "technical": "软件功能错误与故障", "account": "账户密码与身份验证"}}, "completed": {"type": "boolean", "instructions": "读取上述独立记录并回答：退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "读取上述独立记录并回答：按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "documentation", "completed": false, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:documentation:requested:1", "template_id": "support:test:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"billing": 0.01, "technical": 0.02, "documentation": 0.97, "account": 0}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:documentation:processing:0", "state_id": "support:documentation:processing:0", "family_id": "support_decisions_v1", "split": "test", "state": "Independent log — topic: a request for the user manual; refund: the refund is processing and not complete; functionality: all features work without obstruction.", "questions": {"route": {"type": "choice", "instructions": "Read the independent log and answer: Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"shipping": "Parcel transport and delivery", "billing": "Charges, payments, and refunds", "documentation": "Finding manuals and usage guides", "privacy": "Personal data deletion and privacy", "technical": "Software errors and broken features"}}, "completed": {"type": "boolean", "instructions": "Read the independent log and answer: Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Read the independent log and answer: Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "documentation", "completed": false, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:documentation:processing:0", "template_id": "support:test:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"privacy": 0, "shipping": 0, "technical": 0, "documentation": 1, "billing": 0}, "completed": {"false": 0.96, "true": 0.04}, "severity": {"0": 1, "1": 0, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:billing:denied:1", "state_id": "support:billing:denied:1", "family_id": "support_decisions_v1", "split": "test", "state": "请看独立记录——问题：重复扣款；退款：退款申请已拒绝，未付款；功能：次要功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "读取上述独立记录并回答：只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"shipping": "包裹运输与投递问题", "billing": "扣费、付款与退款问题"}}, "completed": {"type": "boolean", "instructions": "读取上述独立记录并回答：退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "读取上述独立记录并回答：按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "billing", "completed": false, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:billing:denied:1", "template_id": "support:test:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"shipping": 0, "billing": 1}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 1, "2": 0, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:account:none:1", "state_id": "support:account:none:1", "family_id": "support_decisions_v1", "split": "test", "state": "请看独立记录——问题：忘记账户密码；退款：尚未提出退款申请；功能：次要功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "读取上述独立记录并回答：只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"account": "账户密码与身份验证", "technical": "软件功能错误与故障"}}, "completed": {"type": "boolean", "instructions": "读取上述独立记录并回答：退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "读取上述独立记录并回答：按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "account", "completed": false, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:account:none:1", "template_id": "support:test:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"account": 0.99, "technical": 0.01}, "completed": {"false": 0.98, "true": 0.02}, "severity": {"0": 0, "1": 0.99, "2": 0.01, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:billing:requested:3", "state_id": "support:billing:requested:3", "family_id": "support_decisions_v1", "split": "test", "state": "Independent log — topic: a duplicate charge; refund: a refund was requested but not processed; functionality: the entire service is down without a workaround.", "questions": {"route": {"type": "choice", "instructions": "Read the independent log and answer: Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"shipping": "Parcel transport and delivery", "billing": "Charges, payments, and refunds", "documentation": "Finding manuals and usage guides"}}, "completed": {"type": "boolean", "instructions": "Read the independent log and answer: Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Read the independent log and answer: Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "billing", "completed": false, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:billing:requested:3", "template_id": "support:test:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"shipping": 0, "billing": 1, "documentation": 0}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:account:processing:3", "state_id": "support:account:processing:3", "family_id": "support_decisions_v1", "split": "test", "state": "Independent log — topic: a forgotten account password; refund: the refund is processing and not complete; functionality: the entire service is down without a workaround.", "questions": {"route": {"type": "choice", "instructions": "Read the independent log and answer: Choose a support team from the main topic only; refund and impact are separate fields.", "criteria": {"documentation": "Finding manuals and usage guides", "billing": "Charges, payments, and refunds", "shipping": "Parcel transport and delivery", "technical": "Software errors and broken features", "account": "Passwords and identity checks"}}, "completed": {"type": "boolean", "instructions": "Read the independent log and answer: Has the refund actually been paid and completed? Requests, processing, approval, or denial do not count."}, "severity": {"type": "score", "instructions": "Read the independent log and answer: Choose the severity level matching the explicitly recorded impact.", "criteria": ["All features work normally", "Minor feature blocked; workaround exists", "Core feature blocked; workaround exists", "Entire service down; no workaround"]}}, "gold": {"route": "account", "completed": false, "severity": 3}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:account:processing:3", "template_id": "support:test:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"account": 1, "billing": 0, "shipping": 0, "documentation": 0, "technical": 0}, "completed": {"false": 0.96, "true": 0.04}, "severity": {"0": 0, "1": 0, "2": 0, "3": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "support:technical:issued:2", "state_id": "support:technical:issued:2", "family_id": "support_decisions_v1", "split": "test", "state": "请看独立记录——问题：软件按钮报错；退款：退款已到账，处理完成；功能：核心功能受阻，有替代办法。", "questions": {"route": {"type": "choice", "instructions": "读取上述独立记录并回答：只按工单主诉选择支持团队；退款记录和功能影响是另外两个字段。", "criteria": {"privacy": "个人资料删除与隐私请求", "account": "账户密码与身份验证", "billing": "扣费、付款与退款问题", "technical": "软件功能错误与故障"}}, "completed": {"type": "boolean", "instructions": "读取上述独立记录并回答：退款是否已经实际到账完成？申请、处理中、批准或拒绝均不算到账。"}, "severity": {"type": "score", "instructions": "读取上述独立记录并回答：按明确记载的使用影响，选择最匹配的严重程度。", "criteria": ["功能正常，无使用障碍", "次要功能受阻，有替代办法", "核心功能受阻，有替代办法", "服务完全中断，无替代办法"]}}, "gold": {"route": "technical", "completed": true, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "support:technical:issued:2", "template_id": "support:test:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"technical": 0.99, "privacy": 0, "account": 0, "billing": 0.01}, "completed": {"false": 0.06999999999999995, "true": 0.93}, "severity": {"0": 0, "1": 0, "2": 1, "3": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "access:staff:activated:0", "state_id": "access:staff:activated:0", "family_id": "access_evidence_ood_v1", "split": "ood", "state": "权限档案：身份为员工；进度为访问权限已经启用；材料为没有提交材料。", "questions": {"route": {"type": "choice", "instructions": "仅按身份和材料决定：访客一律拒绝；员工须身份证明及授权函齐全才允许，否则要求补材料。忽略当前进度。", "criteria": {"request_documents": "材料不足，要求补齐", "grant": "符合规则，允许开通"}}, "completed": {"type": "boolean", "instructions": "访问权限是否已经实际启用？申请或审批通过均不等于启用。"}, "severity": {"type": "score", "instructions": "只评估材料完整程度，忽略身份和申请进度。", "criteria": ["没有材料", "只有身份证明", "身份证明和授权函齐全"]}}, "gold": {"route": "request_documents", "completed": true, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "access:staff:activated:0", "template_id": "access:ood:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"request_documents": 1, "grant": 0}, "completed": {"false": 0.26, "true": 0.74}, "severity": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "access:visitor:approved:2", "state_id": "access:visitor:approved:2", "family_id": "access_evidence_ood_v1", "split": "ood", "state": "Access file: role=visitor; status=approved but not activated; documents=identity proof and authorization letter both provided.", "questions": {"route": {"type": "choice", "instructions": "Decide from role and documents only: deny visitors; allow employees with both identity proof and authorization; otherwise request missing documents. Ignore current status.", "criteria": {"billing": "Route to refund handling", "grant": "Eligible; allow activation", "deny": "Role ineligible; deny activation"}}, "completed": {"type": "boolean", "instructions": "Has access actually been activated? A request or approval does not mean activation."}, "severity": {"type": "score", "instructions": "Assess document completeness only, ignoring role and request status.", "criteria": ["No documents", "Identity proof only", "Identity proof plus authorization"]}}, "gold": {"route": "deny", "completed": false, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "access:visitor:approved:2", "template_id": "access:ood:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"grant": 0.01, "deny": 0.99, "billing": 0}, "completed": {"false": 0.94, "true": 0.06}, "severity": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "access:staff:requested:0", "state_id": "access:staff:requested:0", "family_id": "access_evidence_ood_v1", "split": "ood", "state": "权限档案：身份为员工；进度为仅提交申请；材料为没有提交材料。", "questions": {"route": {"type": "choice", "instructions": "仅按身份和材料决定：访客一律拒绝；员工须身份证明及授权函齐全才允许，否则要求补材料。忽略当前进度。", "criteria": {"grant": "符合规则，允许开通", "billing": "转交退款处理", "request_documents": "材料不足，要求补齐", "archive": "不核查规则，直接归档"}}, "completed": {"type": "boolean", "instructions": "访问权限是否已经实际启用？申请或审批通过均不等于启用。"}, "severity": {"type": "score", "instructions": "只评估材料完整程度，忽略身份和申请进度。", "criteria": ["没有材料", "只有身份证明", "身份证明和授权函齐全"]}}, "gold": {"route": "request_documents", "completed": false, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "access:staff:requested:0", "template_id": "access:ood:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"billing": 0, "archive": 0, "grant": 0, "request_documents": 1}, "completed": {"false": 0.96, "true": 0.04}, "severity": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "access:visitor:activated:2", "state_id": "access:visitor:activated:2", "family_id": "access_evidence_ood_v1", "split": "ood", "state": "Access file: role=visitor; status=access has been activated; documents=identity proof and authorization letter both provided.", "questions": {"route": {"type": "choice", "instructions": "Decide from role and documents only: deny visitors; allow employees with both identity proof and authorization; otherwise request missing documents. Ignore current status.", "criteria": {"deny": "Role ineligible; deny activation", "billing": "Route to refund handling", "grant": "Eligible; allow activation", "archive": "Archive without checking the rules", "request_documents": "Missing evidence; request documents"}}, "completed": {"type": "boolean", "instructions": "Has access actually been activated? A request or approval does not mean activation."}, "severity": {"type": "score", "instructions": "Assess document completeness only, ignoring role and request status.", "criteria": ["No documents", "Identity proof only", "Identity proof plus authorization"]}}, "gold": {"route": "deny", "completed": true, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "access:visitor:activated:2", "template_id": "access:ood:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"grant": 0, "request_documents": 0, "deny": 1, "billing": 0, "archive": 0}, "completed": {"false": 0.22999999999999998, "true": 0.77}, "severity": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "access:staff:pending:2", "state_id": "access:staff:pending:2", "family_id": "access_evidence_ood_v1", "split": "ood", "state": "Access file: role=employee; status=approval pending; not activated; documents=identity proof and authorization letter both provided.", "questions": {"route": {"type": "choice", "instructions": "Decide from role and documents only: deny visitors; allow employees with both identity proof and authorization; otherwise request missing documents. Ignore current status.", "criteria": {"grant": "Eligible; allow activation", "deny": "Role ineligible; deny activation", "request_documents": "Missing evidence; request documents"}}, "completed": {"type": "boolean", "instructions": "Has access actually been activated? A request or approval does not mean activation."}, "severity": {"type": "score", "instructions": "Assess document completeness only, ignoring role and request status.", "criteria": ["No documents", "Identity proof only", "Identity proof plus authorization"]}}, "gold": {"route": "grant", "completed": false, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "access:staff:pending:2", "template_id": "access:ood:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"grant": 0.99, "deny": 0.01, "request_documents": 0}, "completed": {"false": 0.95, "true": 0.05}, "severity": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "access:visitor:pending:2", "state_id": "access:visitor:pending:2", "family_id": "access_evidence_ood_v1", "split": "ood", "state": "权限档案：身份为访客；进度为仍待审批，尚未启用；材料为身份证明和授权函齐全。", "questions": {"route": {"type": "choice", "instructions": "仅按身份和材料决定：访客一律拒绝；员工须身份证明及授权函齐全才允许，否则要求补材料。忽略当前进度。", "criteria": {"deny": "身份不符，拒绝开通", "request_documents": "材料不足，要求补齐", "grant": "符合规则，允许开通", "archive": "不核查规则，直接归档"}}, "completed": {"type": "boolean", "instructions": "访问权限是否已经实际启用？申请或审批通过均不等于启用。"}, "severity": {"type": "score", "instructions": "只评估材料完整程度，忽略身份和申请进度。", "criteria": ["没有材料", "只有身份证明", "身份证明和授权函齐全"]}}, "gold": {"route": "deny", "completed": false, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "access:visitor:pending:2", "template_id": "access:ood:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"archive": 0, "request_documents": 0, "deny": 1, "grant": 0}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "access:staff:requested:1", "state_id": "access:staff:requested:1", "family_id": "access_evidence_ood_v1", "split": "ood", "state": "Access file: role=employee; status=request submitted only; documents=identity proof only.", "questions": {"route": {"type": "choice", "instructions": "Decide from role and documents only: deny visitors; allow employees with both identity proof and authorization; otherwise request missing documents. Ignore current status.", "criteria": {"deny": "Role ineligible; deny activation", "grant": "Eligible; allow activation", "request_documents": "Missing evidence; request documents", "archive": "Archive without checking the rules", "billing": "Route to refund handling"}}, "completed": {"type": "boolean", "instructions": "Has access actually been activated? A request or approval does not mean activation."}, "severity": {"type": "score", "instructions": "Assess document completeness only, ignoring role and request status.", "criteria": ["No documents", "Identity proof only", "Identity proof plus authorization"]}}, "gold": {"route": "request_documents", "completed": false, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "access:staff:requested:1", "template_id": "access:ood:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"billing": 0, "deny": 0, "archive": 0, "request_documents": 0.87, "grant": 0.13}, "completed": {"false": 0.9299999999999999, "true": 0.07}, "severity": {"0": 0, "1": 1, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "access:staff:activated:2", "state_id": "access:staff:activated:2", "family_id": "access_evidence_ood_v1", "split": "ood", "state": "权限档案：身份为员工；进度为访问权限已经启用；材料为身份证明和授权函齐全。", "questions": {"route": {"type": "choice", "instructions": "仅按身份和材料决定：访客一律拒绝；员工须身份证明及授权函齐全才允许，否则要求补材料。忽略当前进度。", "criteria": {"request_documents": "材料不足，要求补齐", "grant": "符合规则，允许开通"}}, "completed": {"type": "boolean", "instructions": "访问权限是否已经实际启用？申请或审批通过均不等于启用。"}, "severity": {"type": "score", "instructions": "只评估材料完整程度，忽略身份和申请进度。", "criteria": ["没有材料", "只有身份证明", "身份证明和授权函齐全"]}}, "gold": {"route": "grant", "completed": true, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "access:staff:activated:2", "template_id": "access:ood:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"request_documents": 0, "grant": 1}, "completed": {"false": 0.16000000000000003, "true": 0.84}, "severity": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "access:visitor:pending:1", "state_id": "access:visitor:pending:1", "family_id": "access_evidence_ood_v1", "split": "ood", "state": "权限档案：身份为访客；进度为仍待审批，尚未启用；材料为仅提交身份证明。", "questions": {"route": {"type": "choice", "instructions": "仅按身份和材料决定：访客一律拒绝；员工须身份证明及授权函齐全才允许，否则要求补材料。忽略当前进度。", "criteria": {"deny": "身份不符，拒绝开通", "grant": "符合规则，允许开通"}}, "completed": {"type": "boolean", "instructions": "访问权限是否已经实际启用？申请或审批通过均不等于启用。"}, "severity": {"type": "score", "instructions": "只评估材料完整程度，忽略身份和申请进度。", "criteria": ["没有材料", "只有身份证明", "身份证明和授权函齐全"]}}, "gold": {"route": "deny", "completed": false, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "access:visitor:pending:1", "template_id": "access:ood:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"deny": 1, "grant": 0}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 0, "1": 1, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "access:visitor:activated:1", "state_id": "access:visitor:activated:1", "family_id": "access_evidence_ood_v1", "split": "ood", "state": "Access file: role=visitor; status=access has been activated; documents=identity proof only.", "questions": {"route": {"type": "choice", "instructions": "Decide from role and documents only: deny visitors; allow employees with both identity proof and authorization; otherwise request missing documents. Ignore current status.", "criteria": {"grant": "Eligible; allow activation", "deny": "Role ineligible; deny activation", "archive": "Archive without checking the rules"}}, "completed": {"type": "boolean", "instructions": "Has access actually been activated? A request or approval does not mean activation."}, "severity": {"type": "score", "instructions": "Assess document completeness only, ignoring role and request status.", "criteria": ["No documents", "Identity proof only", "Identity proof plus authorization"]}}, "gold": {"route": "deny", "completed": true, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "access:visitor:activated:1", "template_id": "access:ood:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"deny": 1, "grant": 0, "archive": 0}, "completed": {"false": 0.25, "true": 0.75}, "severity": {"0": 0, "1": 1, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "access:visitor:requested:0", "state_id": "access:visitor:requested:0", "family_id": "access_evidence_ood_v1", "split": "ood", "state": "权限档案：身份为访客；进度为仅提交申请；材料为没有提交材料。", "questions": {"route": {"type": "choice", "instructions": "仅按身份和材料决定：访客一律拒绝；员工须身份证明及授权函齐全才允许，否则要求补材料。忽略当前进度。", "criteria": {"archive": "不核查规则，直接归档", "grant": "符合规则，允许开通", "billing": "转交退款处理", "deny": "身份不符，拒绝开通"}}, "completed": {"type": "boolean", "instructions": "访问权限是否已经实际启用？申请或审批通过均不等于启用。"}, "severity": {"type": "score", "instructions": "只评估材料完整程度，忽略身份和申请进度。", "criteria": ["没有材料", "只有身份证明", "身份证明和授权函齐全"]}}, "gold": {"route": "deny", "completed": false, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "access:visitor:requested:0", "template_id": "access:ood:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"grant": 0, "deny": 1, "billing": 0, "archive": 0}, "completed": {"false": 0.97, "true": 0.03}, "severity": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "access:staff:approved:1", "state_id": "access:staff:approved:1", "family_id": "access_evidence_ood_v1", "split": "ood", "state": "权限档案：身份为员工；进度为审批通过但尚未启用；材料为仅提交身份证明。", "questions": {"route": {"type": "choice", "instructions": "仅按身份和材料决定：访客一律拒绝；员工须身份证明及授权函齐全才允许，否则要求补材料。忽略当前进度。", "criteria": {"request_documents": "材料不足，要求补齐", "grant": "符合规则，允许开通"}}, "completed": {"type": "boolean", "instructions": "访问权限是否已经实际启用？申请或审批通过均不等于启用。"}, "severity": {"type": "score", "instructions": "只评估材料完整程度，忽略身份和申请进度。", "criteria": ["没有材料", "只有身份证明", "身份证明和授权函齐全"]}}, "gold": {"route": "request_documents", "completed": false, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "access:staff:approved:1", "template_id": "access:ood:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"request_documents": 1, "grant": 0}, "completed": {"false": 0.96, "true": 0.04}, "severity": {"0": 0, "1": 1, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "access:visitor:requested:1", "state_id": "access:visitor:requested:1", "family_id": "access_evidence_ood_v1", "split": "ood", "state": "Access file: role=visitor; status=request submitted only; documents=identity proof only.", "questions": {"route": {"type": "choice", "instructions": "Decide from role and documents only: deny visitors; allow employees with both identity proof and authorization; otherwise request missing documents. Ignore current status.", "criteria": {"billing": "Route to refund handling", "request_documents": "Missing evidence; request documents", "grant": "Eligible; allow activation", "archive": "Archive without checking the rules", "deny": "Role ineligible; deny activation"}}, "completed": {"type": "boolean", "instructions": "Has access actually been activated? A request or approval does not mean activation."}, "severity": {"type": "score", "instructions": "Assess document completeness only, ignoring role and request status.", "criteria": ["No documents", "Identity proof only", "Identity proof plus authorization"]}}, "gold": {"route": "deny", "completed": false, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "access:visitor:requested:1", "template_id": "access:ood:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"grant": 0, "request_documents": 0.02, "archive": 0, "deny": 0.98, "billing": 0}, "completed": {"false": 0.91, "true": 0.09}, "severity": {"0": 0, "1": 1, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "access:visitor:pending:0", "state_id": "access:visitor:pending:0", "family_id": "access_evidence_ood_v1", "split": "ood", "state": "Access file: role=visitor; status=approval pending; not activated; documents=no documents provided.", "questions": {"route": {"type": "choice", "instructions": "Decide from role and documents only: deny visitors; allow employees with both identity proof and authorization; otherwise request missing documents. Ignore current status.", "criteria": {"request_documents": "Missing evidence; request documents", "deny": "Role ineligible; deny activation", "grant": "Eligible; allow activation"}}, "completed": {"type": "boolean", "instructions": "Has access actually been activated? A request or approval does not mean activation."}, "severity": {"type": "score", "instructions": "Assess document completeness only, ignoring role and request status.", "criteria": ["No documents", "Identity proof only", "Identity proof plus authorization"]}}, "gold": {"route": "deny", "completed": false, "severity": 0}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "access:visitor:pending:0", "template_id": "access:ood:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"deny": 0.84, "request_documents": 0.16, "grant": 0}, "completed": {"false": 0.96, "true": 0.04}, "severity": {"0": 1, "1": 0, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "access:staff:requested:2", "state_id": "access:staff:requested:2", "family_id": "access_evidence_ood_v1", "split": "ood", "state": "权限档案：身份为员工；进度为仅提交申请；材料为身份证明和授权函齐全。", "questions": {"route": {"type": "choice", "instructions": "仅按身份和材料决定：访客一律拒绝；员工须身份证明及授权函齐全才允许，否则要求补材料。忽略当前进度。", "criteria": {"grant": "符合规则，允许开通", "deny": "身份不符，拒绝开通", "request_documents": "材料不足，要求补齐", "billing": "转交退款处理"}}, "completed": {"type": "boolean", "instructions": "访问权限是否已经实际启用？申请或审批通过均不等于启用。"}, "severity": {"type": "score", "instructions": "只评估材料完整程度，忽略身份和申请进度。", "criteria": ["没有材料", "只有身份证明", "身份证明和授权函齐全"]}}, "gold": {"route": "grant", "completed": false, "severity": 2}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "access:staff:requested:2", "template_id": "access:ood:zh", "language": "zh"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"deny": 0, "grant": 1, "request_documents": 0, "billing": 0}, "completed": {"false": 0.95, "true": 0.05}, "severity": {"0": 0, "1": 0, "2": 1}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}
{"id": "access:visitor:approved:1", "state_id": "access:visitor:approved:1", "family_id": "access_evidence_ood_v1", "split": "ood", "state": "Access file: role=visitor; status=approved but not activated; documents=identity proof only.", "questions": {"route": {"type": "choice", "instructions": "Decide from role and documents only: deny visitors; allow employees with both identity proof and authorization; otherwise request missing documents. Ignore current status.", "criteria": {"grant": "Eligible; allow activation", "archive": "Archive without checking the rules", "request_documents": "Missing evidence; request documents", "billing": "Route to refund handling", "deny": "Role ineligible; deny activation"}}, "completed": {"type": "boolean", "instructions": "Has access actually been activated? A request or approval does not mean activation."}, "severity": {"type": "score", "instructions": "Assess document completeness only, ignoring role and request status.", "criteria": ["No documents", "Identity proof only", "Identity proof plus authorization"]}}, "gold": {"route": "deny", "completed": false, "severity": 1}, "metadata": {"source": "self_authored_programmatic", "license": "CC0-1.0", "source_group_id": "access:visitor:approved:1", "template_id": "access:ood:en", "language": "en"}, "teacher": {"model": "typesafe-ai/jev", "label_source": "jev_native_evaluation", "target_kind": "rounded_probabilities", "native_probs": {"route": {"deny": 0.91, "archive": 0, "request_documents": 0.09, "grant": 0, "billing": 0}, "completed": {"false": 0.9299999999999999, "true": 0.07}, "severity": {"0": 0, "1": 1, "2": 0}}, "rounding": {"probabilityDecimals": 2, "scoreDecimals": 2}}}