{
  "schema_version": 1,
  "product": "SHAFU 傻福",
  "benchmark": "LiveBench Math",
  "benchmark_release_date": "2024-11-25",
  "evaluation_date": "2026-11-11",
  "model": {
    "label": "SHAFU · 9B 本地增强版",
    "base_model": "Qwen3.5-9B",
    "quantization": "Q4_K_M",
    "parameter_billions": 9,
    "context_tokens": 131072
  },
  "results": {
    "question_count": 232,
    "answered_count": 232,
    "unsubmitted_count": 0,
    "task_mean_percent": 80.56527787979786,
    "task_mean_display": "80.57%",
    "question_mean_percent": 85.46142246128521,
    "question_mean_display": "85.46%",
    "full_credit_question_count": 181,
    "per_task": [
      {
        "task": "olympiad",
        "question_count": 36,
        "mean_percent": 61.86250030606024
      },
      {
        "task": "AMPS_Hard",
        "question_count": 100,
        "mean_percent": 84.0
      },
      {
        "task": "math_comp",
        "question_count": 96,
        "mean_percent": 95.83333333333333
      }
    ]
  },
  "methodology": {
    "label": "内部开发评测",
    "headline_note": "多轮迭代汇总，非官方榜单成绩。",
    "task_mean_definition": "分别计算 olympiad、AMPS_Hard、math_comp 的平均分，再对三个任务等权平均。",
    "question_mean_definition": "232 道题按题等权平均；其中部分题目允许部分得分。",
    "full_credit_definition": "181 表示得到满分的单题数量，不表示整个基准为 100%。",
    "evaluation_scope": "增强系统的内部开发记录，汇总多轮不同版本的有效结果；不是一次冻结配置的完整独立测试。",
    "caveat": "题目曾用于开发与迭代；结果不代表未见题目的表现，不能直接视为基础模型单独能力或官方排行榜提交成绩。",
    "third_party_rank_claim": false,
    "single_frozen_run": false
  },
  "full_credit_examples": [
    {
      "label": "Q01",
      "scope": "单题开发案例",
      "earlier_score_percent": 71.42857142857143,
      "score_percent": 100.0,
      "note": "历史单题案例，不是独立测试集的总体准确率。"
    },
    {
      "label": "Q03",
      "scope": "单题开发案例",
      "earlier_score_percent": 100.0,
      "score_percent": 100.0,
      "note": "历史单题案例，不是独立测试集的总体准确率。"
    }
  ]
}