{
  "description": "Released training prompt sets of DN-MOPD. Check or rebuild with data/build/build_data.py.",
  "row_schema": {
    "prompt": "chat messages [{role, content}] (student, math, code) or a plain string (teacher IF; the trainer wraps it as one user message)",
    "label": "math: reference answer; code: JSON string {inputs, outputs} of test cases; IF: JSON string list of constraints",
    "domain": "math | code | ifeval",
    "teacher": "teacher routing name (= domain)",
    "metadata": "{domain, data_source, src_index}: domain selects the verifier; src_index is the row's position in its source pool"
  },
  "files": {
    "student/train_3domain.jsonl": {
      "sha256": "5317739721646bd5394775eaef0019e4fded5ac5b8d481e61a2ba8bd754ac08e",
      "bytes": 45968876,
      "rows": 2700,
      "domains": {
        "code": 900,
        "ifeval": 900,
        "math": 900
      },
      "data_sources": {
        "DeepMath-103K": 900,
        "apps": 156,
        "codecontests": 370,
        "codeforces": 92,
        "synthetic-ifeval-v1": 900,
        "taco": 282
      },
      "in_git": true,
      "used_by": "every student run (Label, DN-MOPD, single-teacher OPD, uniform pool, dynamic router, annealed injection, SeqKD-SFT generation, controls)"
    },
    "teacher/math_train.jsonl": {
      "sha256": "60ef3d6dee3b405a81ed51d8b646e403228bd6f49e3412a987f90783da4e331c",
      "bytes": 3838298,
      "rows": 8000,
      "domains": {
        "math": 8000
      },
      "data_sources": {
        "DeepMath-103K": 8000
      },
      "in_git": true,
      "used_by": "GRPO math teachers"
    },
    "teacher/code_train.jsonl": {
      "sha256": "f52b5143f0ed1721e1bba3d997e432151794a09e03b60c045ecaa3c764d5c9e8",
      "bytes": 917096848,
      "rows": 8000,
      "domains": {
        "code": 8000
      },
      "data_sources": {
        "apps": 1087,
        "codecontests": 3069,
        "codeforces": 853,
        "taco": 2991
      },
      "in_git": false,
      "used_by": "GRPO code teachers"
    },
    "teacher/if_train.jsonl": {
      "sha256": "60b169a376ed9fffc445b1ffb8a7adfe81d7c9ffacc7b253324385b2367d96d8",
      "bytes": 3722418,
      "rows": 8000,
      "domains": {
        "ifeval": 8000
      },
      "data_sources": {
        "synthetic-ifeval-v1": 8000
      },
      "in_git": true,
      "used_by": "GRPO IF teachers"
    }
  }
}