README: sharper intro, task questions, build guide

README.md

@@ -10,9 +10,14 @@ task_categories:
 tags:
 - agents
 - tool-calling
+- tool-use
+- function-calling
+- decision-models
+- routing
 - structured-prediction
 - preference-data
-- routing
+size_categories:
+- 100K<n<1M
 configs:
 - config_name: default
   data_files:
@@ -26,96 +31,91 @@ configs:
 
 # Agent Tool Decisions 180K
 
-**180,000 validated, source-labeled typed decisions for modern AI
-systems.** Typed decisions for whether an agent should call a tool, which tool to call, whether arguments are complete, and whether a tool-using response is preferable.
+**180,000 typed decisions that sit behind every agent step.** Should the agent call a tool, or
+answer in text? Which tool? Are the arguments complete? Which tool-using response is better?
 
-> **General-purpose dataset:** this release contains no clinical or medical
-> data and no Jev or TypeSafe output columns.
+An agent spends most of its time making these small decisions, and a frontier LLM is an
+expensive way to make them. They are **typed**: a question, a fixed set of choices, one correct
+answer. That makes them a good fit for small, fast decision models (encoders, classifiers,
+Jev-style typed decision models, GLiNER2.5-Decide-style heads) that answer in milliseconds on a
+CPU.
 
-Each row preserves an upstream human, automated, or programmatic label plus
-the typed state, decision question, criteria, split identity, and pinned source
-provenance. The flat Parquet layout works directly with Hugging Face
-`datasets`, DuckDB, Polars, or pandas.
+This dataset turns NVIDIA's open agentic data into exactly that format: 180K clean, labeled,
+split-safe decision rows, ready to train or evaluate a tool router today.
 
-## What it is good for
+> **Built on NVIDIA's open data:** `nvidia/Nemotron-SFT-Agentic-v2` and `nvidia/When2Call`.
+> Every label comes from the upstream data (programmatic or automated). **No output from Jev,
+> TypeSafe or any other proprietary model is included.** It contains no clinical or medical
+> data.
 
-- tool-use routers and next-action classifiers
-- tool-selection and argument-completeness heads
-- agent policy evaluation and selective prediction
-- reward/preference modeling for tool-using responses
-- benchmarking and error analysis for typed agent decisions
+## At a glance
 
-## Release summary
-
-| Item | Value |
+| | |
 | --- | ---: |
-| Rows / unique row IDs | 180,000 |
-| Unique decision groups | 101,786 |
-| Unique upstream source rows | 41,809 |
-| Train | 171,056 |
-| Validation | 2,713 |
-| Test | 6,231 |
-| Columns | 35 |
-| Jev/TypeSafe outputs | None |
+| Rows | **180,000** |
+| Train / validation / test | 171,056 / 2,713 / 6,231 |
+| Decision groups | 101,786 |
+| Upstream source rows | 41,809 |
+| Leakage | none: no decision group or source row crosses splits |
+| Format | flat Parquet, 35 columns, works with `datasets`, DuckDB, Polars and pandas |
+
+## The six decisions
+
+| Task family | Question the model answers | Rows |
+| --- | --- | ---: |
+| `agent_next_action_type` | What type of action should the assistant take next? | 79,238 |
+| `tool_selection` | Which available tool should be called next? | 37,233 |
+| `tool_argument_completeness` | Do the proposed tool calls include every required argument named by the tool schemas? | 37,233 |
+| `tool_or_text_action` | Should the assistant call a tool or produce a text response next? | 14,349 |
+| `tool_response_preference` | Which response better uses or declines the available tools? | 8,295 |
+| `when_to_call_tool` | When2Call-style: should a tool be called at all? | 3,652 |
+
+Two answer types (`primitive`): `choice` (one label from the criteria, 142,767 rows) and
+`noul` (a bounded 0–1 value, 37,233 rows).
 
 ## Quick start
 
 ```python
 from datasets import load_dataset
 
-dataset = load_dataset("MaziyarPanahi/AgentToolDecisions-180K")
-row = dataset["train"][0]
+ds = load_dataset("MaziyarPanahi/AgentToolDecisions-180K")
+row = ds["train"][0]
+
+print(row["question_text"])   # e.g. "Which available tool should be called next?"
+print(row["state_preview"])   # the agent state: conversation + available tools
+print(row["criteria_json"])   # the allowed answers
+print(row["gold_label"] if row["gold_label"] is not None else row["gold_score"])
+```
+
+Train a tool router on a single task family:
 
-print(row["state_preview"])
-print(row["question_text"])
-print(row["gold_label"] or row["gold_score"])
+```python
+tool_sel = ds["train"].filter(lambda r: r["task_family"] == "tool_selection")
+# inputs: question_text + state_json, labels: gold_label (one of the keys in criteria_json)
 ```
 
-For a private repository, authenticate first with `hf auth login`. Once the
-repository is public, the same snippet works without authentication.
+## What you can build with it
+
+- **Tool routers:** pick the next tool in milliseconds instead of spending a full LLM call.
+- **"Should I call a tool?" gates:** cut useless tool calls and the latency they add.
+- **Argument checkers:** catch missing required arguments before the call fails.
+- **Reward and preference models** for tool-using responses.
+- **Evaluation:** measure your agent's decision accuracy per task family, on a leakage-free test
+  split.
 
 ## How to read a row
 
 | Need | Columns |
 | --- | --- |
-| Stable identity and split | `row_id`, `group_id`, `output_split` |
-| Decision task | `priority`, `task_family`, `primitive` |
-| Model input | `state_json`, `state_preview`, `question_text`, `criteria_json` |
+| Identity and split | `row_id`, `group_id`, `output_split` |
+| Decision task | `priority`, `task_family`, `primitive`, `question_key` |
+| Model input | `state_json`, `state_preview`, `question_text`, `criteria_json`, `request_json` |
 | Reference answer | `gold_label`, `gold_score`, `gold_json`, `label_source` |
-| Provenance | source dataset, pinned revision, row ID, license, URL, transformation |
-
-Decision primitives are intentionally simple:
-
-- `choice`: one categorical label;
-- `noul`: a bounded value from 0 to 1;
-- `score`: a structured numeric assessment.
-
-## Task mix
-
-| Task family | Rows |
-| --- | ---: |
-| `agent_next_action_type` | 79,238 |
-| `tool_argument_completeness` | 37,233 |
-| `tool_selection` | 37,233 |
-| `tool_or_text_action` | 14,349 |
-| `tool_response_preference` | 8,295 |
-| `when_to_call_tool` | 3,652 |
-
-## Decision primitives
-
-| Primitive | Rows |
-| --- | ---: |
-| `choice` | 142,767 |
-| `noul` | 37,233 |
+| Provenance | `source_dataset`, `source_revision`, `source_row_id`, `source_license`, `source_url`, `transformation`, `transform_version` |
 
-## Source mix
+Nested payloads are canonical JSON strings, so the Parquet schema stays stable across tasks.
 
-| Source | Rows |
-| --- | ---: |
-| `nvidia/Nemotron-SFT-Agentic-v2` | 153,704 |
-| `nvidia/When2Call` | 26,296 |
-
-## Label provenance
+## Where the labels come from
 
 | Label source | Rows |
 | --- | ---: |
@@ -126,42 +126,33 @@ Decision primitives are intentionally simple:
 | `automated_preference_gold` | 8,295 |
 | `automated_verified_mcq_gold` | 3,652 |
 
-Every row retains the pinned source revision, source license, source row ID,
-transformation, group ID, original label, typed state, question, criteria, and
-complete structured request. Nested payloads are canonical JSON strings so the
-Parquet schema is stable across tasks.
-
-## Validation and splitting
+| Source | Rows |
+| --- | ---: |
+| `nvidia/Nemotron-SFT-Agentic-v2` | 153,704 |
+| `nvidia/When2Call` | 26,296 |
 
-The release uses repaired component-safe splits. Its independent audit verifies:
+## Validation
 
-- exactly 180,000 rows and unique row IDs;
-- one priority only: `tool_action_gate`;
+An independent audit checks:
+- 180,000 rows with unique row IDs;
 - no decision group or upstream source row crosses splits;
-- valid canonical state, request, criteria, gold, and metadata JSON;
-- choice labels inside declared criteria and bounded binary scores;
-- nonempty pinned source revisions, licenses, URLs, and transformations;
-- no Jev prediction, answer, probability, raw-response, usage, or latency columns;
-- flat, viewer-friendly Parquet shards.
-
-## Important limitations
-
-- Tool schemas and domains vary across upstream agent episodes.
-- Upstream labels can be noisy or ambiguous; retain label provenance and task metadata together.
-- Upstream labels can be noisy, ambiguous, automated, or task-specific; they are not universal ground truth.
-- Tool schemas and domain taxonomies vary across episodes.
-- The typed request contract is a normalized transformation, not the original upstream schema.
-- This is a general decision-modeling dataset, not a clinical dataset.
-
-## Sources and licensing
-
-Rows come from pinned revisions of:
-
-- `nvidia/Nemotron-SFT-Agentic-v2`
-- `nvidia/When2Call`
-
-The When2Call source is marked CC BY 4.0. The Nemotron repository card lists
-CC BY 4.0, Apache 2.0, and MIT terms across its components. Every row retains
-its recorded source license and provenance; users must review and comply with
-the applicable upstream dataset cards and terms. No Jev output is included,
-so this repository is distinct from the Jev-enriched edition.
+- valid canonical JSON for state, request, criteria, gold and metadata;
+- choice labels inside the declared criteria, and scores within bounds;
+- a pinned source revision, license, URL and transformation on every row;
+- no proprietary-model prediction, probability, raw-response, usage or latency columns.
+
+## Limitations
+
+- Labels are programmatic or automated upstream labels. They can be noisy or ambiguous, so keep
+  `label_source` and `task_family` with every prediction you evaluate.
+- Tool schemas and domains vary across the upstream agent episodes.
+- The typed request format is a normalized transformation, not the original upstream schema.
+- This is a general agent-decision dataset, not a clinical one.
+
+## License and attribution
+
+Derived from pinned revisions of `nvidia/Nemotron-SFT-Agentic-v2` and `nvidia/When2Call`. When2Call
+is marked CC BY 4.0. The Nemotron card lists CC BY 4.0, Apache 2.0 and MIT terms across its
+components. Every row keeps its recorded source license. Please review and follow the upstream
+dataset cards when you use or redistribute the data. Thanks to NVIDIA for releasing the source data
+openly.