{
  "checkpoint": {
    "repo_id": "AustinFu/Qev-9B",
    "revision": "v0.2.0",
    "base": "Qwen/Qwen3.5-9B-Base",
    "base_revision": "68c46c4b3498877f3ef123c856ecfde50c39f404",
    "step": 2643,
    "dtype": "bf16",
    "execution": "reference",
    "seed": 17,
    "temperature": 1.0
  },
  "config": "configs/qev-9b.json",
  "training_recipe": {
    "main_records": 39605,
    "main_questions": 45887,
    "principle_records": 4459,
    "boundary_records": 600,
    "boundary_partition": "main",
    "late_records": 1783,
    "late_fraction": 0.5,
    "late_repeats": 3,
    "epochs": 2,
    "global_batch": 32
  },
  "protocol": {
    "execution": "full causal reference",
    "metric": "argmax accuracy; rejected questions count as wrong",
    "jevbench_version": "v1.4.2",
    "jevbench_upstream_revision": "1df665e3956d7aab7fa0208ff6c4f2d8557f9f90",
    "trained_limits": {
      "max_state": 1024,
      "max_question": 512,
      "max_candidate": 256,
      "max_path": 2048,
      "max_candidates": 128
    }
  },
  "results": {
    "decision_dev_all": {
      "correct": 1295,
      "total": 1468
    },
    "decision_dev_clean": {
      "correct": 1105,
      "total": 1264
    },
    "transfer_dev_all": {
      "correct": 630,
      "total": 764
    },
    "transfer_dev_clean": {
      "correct": 551,
      "total": 656
    },
    "mmlupro": {
      "correct": 574,
      "total": 1000
    },
    "semif_handwritten": {
      "correct": 134,
      "total": 144
    },
    "scienthoon": {
      "correct": 620,
      "total": 873
    },
    "wanli": {
      "correct": 182,
      "total": 256
    },
    "jevbench_public": {
      "correct": 192,
      "total": 231
    }
  }
}