{
  "model": "Qev-4B",
  "repo_id": "AustinFu/Qev-4B",
  "revision": "v0.1.0",
  "base": "Qwen/Qwen3.5-4B-Base",
  "base_revision": "710fd005d44d55ee27b7ad5147e318e546efdbfe",
  "step": 2786,
  "initialization": "Qwen base with fresh LoRA and decision head",
  "training": {
    "questions": 44576,
    "epochs": 2,
    "global_batch": 32,
    "world_size": 4,
    "seed": 17,
    "objective": "teacher-probability cross entropy",
    "teacher_temperature": 1.563437713227029,
    "student_temperature": 1.0,
    "teacher": "AustinFu/Qev-9B@v0.3.0",
    "late_split": null,
    "online_option_augmentation": false
  },
  "evaluation": {
    "execution": "reference",
    "backbone_dtype": "bf16",
    "head_dtype": "fp32",
    "temperature": 1.0,
    "max_state": 4096,
    "max_path": 4096,
    "recorded_at": "2026-10-02",
    "rejected_questions": 0
  },
  "matched_subsets": {
    "decision_dev_all": {
      "correct": 1286,
      "total": 1468
    },
    "decision_dev_clean": {
      "correct": 1099,
      "total": 1264
    },
    "transfer_dev_all": {
      "correct": 618,
      "total": 764
    },
    "transfer_dev_clean": {
      "correct": 538,
      "total": 656
    },
    "mmlupro": {
      "correct": 503,
      "total": 1000
    },
    "scienthoon": {
      "correct": 670,
      "total": 873
    },
    "wanli": {
      "correct": 188,
      "total": 256
    },
    "semif_handwritten": {
      "correct": 130,
      "total": 144
    },
    "jevbench_public": {
      "correct": 190,
      "total": 231
    }
  }
}