source_inventory.json

{
 "dataset": "prism-training-dataset",
 "records": 277496,
 "files": {
  "if_eval.jsonl": {
   "sha256": "8ff20c399072433afa04b13c82358f5f3eff3f784a0215b048233bdcd54417e1",
   "records": 492
  },
  "if_multi_constraints.jsonl": {
   "sha256": "d24ed6fa26852174d79d96d1cf968c8f4e2a0b0c47eef04d59f4061c48669a9f",
   "records": 77002
  },
  "ultrachat.jsonl": {
   "sha256": "f74eef7afb15504996aaad073fb800dd36e1293f1b952e14009f8ef7a1e40248",
   "records": 200002
  },
  "valid_record_ids.json": {
   "sha256": "66202284fffb28e2e51401ddbb4230da916dfda6033c677645de0fa78703314a",
   "ids": 203589
  }
 },
 "generation": {
  "target_model": "Qwen/Qwen3.5-9B (apache-2.0); response = its reply to prompt (max 2048 tokens, temperature sampling); instruction_set = its instruction listing for prompt (max 512 tokens)",
  "filter": "rules + LLM-judge mask (valid_record_ids.json, 203589 ids); split seed 42, group-aware, 0.8/0.1/0.1"
 },
 "sources": [
  {
   "name": "IFEval",
   "hf_id": "google/IFEval",
   "revision": "966cd89545d6",
   "license": "apache-2.0",
   "included_fields": "prompts (verbatim and paraphrased)",
   "transformations": "prompt selection; paraphrasing by the target model",
   "redistribution": "permitted with attribution (Apache-2.0)"
  },
  {
   "name": "IF_multi_constraints_upto5",
   "hf_id": "allenai/IF_multi_constraints_upto5",
   "revision": "2e3a77407b7f",
   "license": "odc-by",
   "included_fields": "prompts (verbatim and paraphrased)",
   "transformations": "prompt selection; paraphrasing by the target model",
   "redistribution": "permitted with attribution (ODC-BY 1.0; attribute AllenAI)"
  },
  {
   "name": "ultrachat_200k",
   "hf_id": "HuggingFaceH4/ultrachat_200k",
   "revision": "8049631c405a",
   "license": "mit",
   "included_fields": "first-turn prompts (verbatim and paraphrased)",
   "transformations": "prompt selection; paraphrasing by the target model",
   "redistribution": "permitted with attribution (MIT)"
  }
 ]
}