{
  "at": "2026-09-20T06:54:32.250982+00:00",
  "model": "sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2",
  "model_snapshot": "e8f8c211226b894fcb81acc59f3b34ba3efd5f42",
  "device": "CPU",
  "dimensions": 384,
  "chunks": 31,
  "index_encode_ms": 1160.17,
  "batch_query_encode_ms": 39.64,
  "runtime_seconds": 5.13,
  "summary": {
    "baseline": {
      "top1": 17,
      "top4": 21,
      "total": 21
    },
    "v2": {
      "top1": 19,
      "top4": 21,
      "total": 21
    },
    "neural": {
      "top1": 19,
      "top4": 21,
      "total": 21
    }
  },
  "retrieval_cases": [
    {
      "query": "新员工报到地点",
      "gold": "01-1",
      "baseline_rank": 1,
      "v2_rank": 1,
      "v2_ids": [
        "01-1",
        "01-3",
        "01-5",
        "01-2"
      ],
      "neural_ids": [
        "01-1",
        "01-3",
        "01-4",
        "01-5"
      ],
      "neural_scores": [
        0.64891,
        0.56354,
        0.54103,
        0.516
      ],
      "neural_rank": 1
    },
    {
      "query": "入职需要带哪些材料",
      "gold": "01-2",
      "baseline_rank": 1,
      "v2_rank": 1,
      "v2_ids": [
        "01-2",
        "01-4",
        "01-3",
        "05-2"
      ],
      "neural_ids": [
        "01-2",
        "01-3",
        "04-3",
        "03-4"
      ],
      "neural_scores": [
        0.67954,
        0.48259,
        0.43622,
        0.41056
      ],
      "neural_rank": 1
    },
    {
      "query": "入职材料缺失怎么办",
      "gold": "01-4",
      "baseline_rank": 3,
      "v2_rank": 2,
      "v2_ids": [
        "04-5",
        "01-4",
        "01-3",
        "01-2"
      ],
      "neural_ids": [
        "01-2",
        "01-4",
        "01-3",
        "04-5"
      ],
      "neural_scores": [
        0.59681,
        0.52985,
        0.46801,
        0.44928
      ],
      "neural_rank": 2
    },
    {
      "query": "事假提前多久申请",
      "gold": "02-2",
      "baseline_rank": 1,
      "v2_rank": 1,
      "v2_ids": [
        "02-2",
        "04-5",
        "02-1",
        "04-2"
      ],
      "neural_ids": [
        "02-2",
        "01-4",
        "02-4",
        "04-2"
      ],
      "neural_scores": [
        0.67332,
        0.58103,
        0.57397,
        0.56678
      ],
      "neural_rank": 1
    },
    {
      "query": "病假需要医疗证明吗",
      "gold": "02-3",
      "baseline_rank": 1,
      "v2_rank": 1,
      "v2_ids": [
        "02-3",
        "02-2",
        "02-4",
        "01-2"
      ],
      "neural_ids": [
        "02-3",
        "02-1",
        "02-2",
        "05-5"
      ],
      "neural_scores": [
        0.49635,
        0.40021,
        0.33602,
        0.26965
      ],
      "neural_rank": 1
    },
    {
      "query": "紧急请假返岗后补办期限",
      "gold": "02-4",
      "baseline_rank": 1,
      "v2_rank": 1,
      "v2_ids": [
        "02-4",
        "02-6",
        "02-2",
        "01-4"
      ],
      "neural_ids": [
        "02-4",
        "01-1",
        "01-4",
        "02-2"
      ],
      "neural_scores": [
        0.74781,
        0.6368,
        0.57741,
        0.53822
      ],
      "neural_rank": 1
    },
    {
      "query": "漏打卡补卡截止时间",
      "gold": "02-5",
      "baseline_rank": 2,
      "v2_rank": 1,
      "v2_ids": [
        "02-5",
        "02-1",
        "02-2",
        "02-3"
      ],
      "neural_ids": [
        "02-5",
        "01-4",
        "01-1",
        "03-5"
      ],
      "neural_scores": [
        0.63858,
        0.47657,
        0.43666,
        0.42743
      ],
      "neural_rank": 1
    },
    {
      "query": "出差交通费用标准",
      "gold": "03-2",
      "baseline_rank": 1,
      "v2_rank": 1,
      "v2_ids": [
        "03-2",
        "03-7",
        "03-4",
        "03-1"
      ],
      "neural_ids": [
        "03-2",
        "03-4",
        "03-1",
        "03-3"
      ],
      "neural_scores": [
        0.77688,
        0.48705,
        0.46326,
        0.44979
      ],
      "neural_rank": 1
    },
    {
      "query": "住宿费标准",
      "gold": "03-3",
      "baseline_rank": 1,
      "v2_rank": 1,
      "v2_ids": [
        "03-3",
        "03-4",
        "03-6",
        "03-2"
      ],
      "neural_ids": [
        "03-3",
        "03-2",
        "03-4",
        "03-6"
      ],
      "neural_scores": [
        0.70219,
        0.54056,
        0.41889,
        0.35037
      ],
      "neural_rank": 1
    },
    {
      "query": "差旅报销材料提交方式",
      "gold": "03-4",
      "baseline_rank": 1,
      "v2_rank": 1,
      "v2_ids": [
        "03-4",
        "04-3",
        "03-5",
        "01-4"
      ],
      "neural_ids": [
        "03-4",
        "01-2",
        "04-3",
        "03-7"
      ],
      "neural_scores": [
        0.59339,
        0.50301,
        0.4897,
        0.4359
      ],
      "neural_rank": 1
    },
    {
      "query": "报销提交期限与审核",
      "gold": "03-5",
      "baseline_rank": 1,
      "v2_rank": 1,
      "v2_ids": [
        "03-5",
        "03-4",
        "04-5",
        "03-7"
      ],
      "neural_ids": [
        "03-5",
        "04-4",
        "03-4",
        "04-2"
      ],
      "neural_scores": [
        0.67798,
        0.57963,
        0.56211,
        0.54467
      ],
      "neural_rank": 1
    },
    {
      "query": "票据丢失如何报销",
      "gold": "03-6",
      "baseline_rank": 3,
      "v2_rank": 1,
      "v2_ids": [
        "03-6",
        "03-4",
        "03-7",
        "03-2"
      ],
      "neural_ids": [
        "03-6",
        "03-2",
        "03-3",
        "02-5"
      ],
      "neural_scores": [
        0.56285,
        0.40699,
        0.29982,
        0.29024
      ],
      "neural_rank": 1
    },
    {
      "query": "试用期多长",
      "gold": "04-1",
      "baseline_rank": 1,
      "v2_rank": 1,
      "v2_ids": [
        "04-1",
        "04-6",
        "04-5",
        "04-3"
      ],
      "neural_ids": [
        "01-4",
        "03-5",
        "02-2",
        "04-1"
      ],
      "neural_scores": [
        0.37775,
        0.36513,
        0.35469,
        0.33298
      ],
      "neural_rank": 4
    },
    {
      "query": "转正申请材料",
      "gold": "04-3",
      "baseline_rank": 1,
      "v2_rank": 2,
      "v2_ids": [
        "04-5",
        "04-3",
        "04-2",
        "04-6"
      ],
      "neural_ids": [
        "04-3",
        "04-5",
        "01-2",
        "01-3"
      ],
      "neural_scores": [
        0.57684,
        0.51339,
        0.50581,
        0.50389
      ],
      "neural_rank": 1
    },
    {
      "query": "转正评估结果怎么确认",
      "gold": "04-4",
      "baseline_rank": 1,
      "v2_rank": 1,
      "v2_ids": [
        "04-4",
        "04-2",
        "04-6",
        "04-5"
      ],
      "neural_ids": [
        "04-4",
        "04-2",
        "03-5",
        "01-4"
      ],
      "neural_scores": [
        0.5861,
        0.37703,
        0.37576,
        0.36298
      ],
      "neural_rank": 1
    },
    {
      "query": "转正申请材料缺失",
      "gold": "04-5",
      "baseline_rank": 1,
      "v2_rank": 1,
      "v2_ids": [
        "04-5",
        "04-3",
        "01-4",
        "04-2"
      ],
      "neural_ids": [
        "04-5",
        "04-3",
        "01-2",
        "01-4"
      ],
      "neural_scores": [
        0.59844,
        0.46262,
        0.43834,
        0.43198
      ],
      "neural_rank": 1
    },
    {
      "query": "新账号开通处理进度",
      "gold": "05-2",
      "baseline_rank": 1,
      "v2_rank": 1,
      "v2_ids": [
        "05-2",
        "01-2",
        "01-3",
        "05-7"
      ],
      "neural_ids": [
        "05-2",
        "05-3",
        "05-6",
        "01-3"
      ],
      "neural_scores": [
        0.5196,
        0.36126,
        0.3494,
        0.32339
      ],
      "neural_rank": 1
    },
    {
      "query": "登录失败密码重置",
      "gold": "05-3",
      "baseline_rank": 1,
      "v2_rank": 1,
      "v2_ids": [
        "05-3",
        "05-7",
        "05-5",
        "05-6"
      ],
      "neural_ids": [
        "05-3",
        "04-5",
        "05-5",
        "05-6"
      ],
      "neural_scores": [
        0.73487,
        0.42379,
        0.37068,
        0.35733
      ],
      "neural_rank": 1
    },
    {
      "query": "系统权限申请变更",
      "gold": "05-4",
      "baseline_rank": 1,
      "v2_rank": 1,
      "v2_ids": [
        "05-4",
        "05-7",
        "02-2",
        "05-1"
      ],
      "neural_ids": [
        "05-4",
        "04-5",
        "04-2",
        "05-1"
      ],
      "neural_scores": [
        0.71598,
        0.50371,
        0.50304,
        0.48854
      ],
      "neural_rank": 1
    },
    {
      "query": "发现可疑邮件报告给谁",
      "gold": "05-5",
      "baseline_rank": 1,
      "v2_rank": 1,
      "v2_ids": [
        "05-5",
        "05-1",
        "05-4",
        "05-3"
      ],
      "neural_ids": [
        "05-5",
        "05-1",
        "01-5",
        "05-7"
      ],
      "neural_scores": [
        0.55548,
        0.38651,
        0.37444,
        0.3334
      ],
      "neural_rank": 1
    },
    {
      "query": "账号借给同事使用",
      "gold": "05-6",
      "baseline_rank": 3,
      "v2_rank": 1,
      "v2_ids": [
        "05-6",
        "05-1",
        "04-4",
        "05-7"
      ],
      "neural_ids": [
        "05-6",
        "05-2",
        "05-7",
        "05-1"
      ],
      "neural_scores": [
        0.51795,
        0.47,
        0.44853,
        0.44713
      ],
      "neural_rank": 1
    }
  ],
  "checks": [
    {
      "name": "向量库排名与精确余弦排名一致",
      "pass": true,
      "detail": [
        "01-1",
        "01-3",
        "01-4",
        "01-5"
      ]
    },
    {
      "name": "元数据过滤不返回其他文件",
      "pass": true,
      "detail": "01-新员工入职办理指南.md"
    },
    {
      "name": "重新打开持久化集合",
      "pass": true,
      "detail": 31
    },
    {
      "name": "更新不产生重复分块",
      "pass": true,
      "detail": 31
    },
    {
      "name": "算术工具实际执行",
      "pass": true,
      "detail": 42
    },
    {
      "name": "订单不存在的稳定结果",
      "pass": true,
      "detail": "模拟订单数据"
    },
    {
      "name": "未知工具拒绝",
      "pass": true,
      "detail": "本地协议执行检查"
    },
    {
      "name": "多余参数拒绝",
      "pass": true,
      "detail": "本地协议执行检查"
    },
    {
      "name": "数值类型拒绝",
      "pass": true,
      "detail": "本地协议执行检查"
    },
    {
      "name": "重复写入幂等",
      "pass": true,
      "detail": 1
    },
    {
      "name": "幂等键冲突拒绝",
      "pass": true,
      "detail": "同键不同负载不覆盖"
    }
  ],
  "notes": [
    "同一21题开发集，非盲测。向量模型无微调；不据此宣称总体优劣。",
    "Chroma测试采用隔离新集合，不改原项目数据库。",
    "工具协议测试使用本地模拟订单与CRM，没有调用真实业务服务或真实天气。",
    "HR应用当前选择词法检索；这里的神经向量实验单独交付，未伪称已上线到应用。"
  ]
}