{
  "metrics": [
    {
      "id": "hard_pass",
      "mode": "offline",
      "name": "核心硬断言通过率",
      "value": 0.8513513513513513,
      "numerator": 63,
      "denominator": 74,
      "unit": "ratio",
      "target": null,
      "scope": "74个核心任务首轮",
      "status": "MEASURED"
    },
    {
      "id": "pass_1",
      "mode": "offline",
      "name": "pass^1",
      "value": 0.8513513513513513,
      "numerator": 63,
      "denominator": 74,
      "target": null,
      "unit": "ratio",
      "status": "MEASURED",
      "scope": "核心任务前k次全部硬断言通过；不含重复语义质量"
    },
    {
      "id": "pass_3",
      "mode": "offline",
      "name": "pass^3",
      "value": 0.8513513513513513,
      "numerator": 63,
      "denominator": 74,
      "target": 0.85,
      "unit": "ratio",
      "status": "MEASURED",
      "scope": "核心任务前k次全部硬断言通过；不含重复语义质量"
    },
    {
      "id": "pass_8",
      "mode": "offline",
      "name": "pass^8",
      "value": 0.8513513513513513,
      "numerator": 63,
      "denominator": 74,
      "target": 0.6,
      "unit": "ratio",
      "status": "MEASURED",
      "scope": "核心任务前k次全部硬断言通过；不含重复语义质量"
    },
    {
      "id": "order",
      "mode": "offline",
      "name": "订单字段完整且一致",
      "value": 1.0,
      "numerator": 12,
      "denominator": 12,
      "unit": "ratio",
      "target": 1,
      "scope": "首轮对应自动化断言；引用PN按引用条目计，非任务数",
      "status": "MEASURED"
    },
    {
      "id": "pn",
      "mode": "offline",
      "name": "引用PN一致",
      "value": 1.0,
      "numerator": 35,
      "denominator": 35,
      "unit": "ratio",
      "target": 1,
      "scope": "首轮对应自动化断言；引用PN按引用条目计，非任务数",
      "status": "MEASURED"
    },
    {
      "id": "auth",
      "mode": "offline",
      "name": "动作边界遵循",
      "value": 1.0,
      "numerator": 14,
      "denominator": 14,
      "unit": "ratio",
      "target": 1,
      "scope": "首轮对应自动化断言；引用PN按引用条目计，非任务数",
      "status": "MEASURED"
    },
    {
      "id": "robust",
      "mode": "offline",
      "name": "等义扰动终态一致",
      "value": 1.0,
      "numerator": 3,
      "denominator": 3,
      "unit": "ratio",
      "target": 0.8,
      "scope": "首轮对应自动化断言；引用PN按引用条目计，非任务数",
      "status": "MEASURED"
    },
    {
      "id": "handoff",
      "mode": "offline",
      "name": "转人工上下文字段完整",
      "value": 0.0,
      "numerator": 0,
      "denominator": 1,
      "unit": "ratio",
      "target": 1,
      "scope": "首轮对应自动化断言；引用PN按引用条目计，非任务数",
      "status": "MEASURED"
    },
    {
      "id": "restore",
      "mode": "offline",
      "name": "同进程会话恢复",
      "value": 1.0,
      "numerator": 1,
      "denominator": 1,
      "unit": "ratio",
      "target": 0.9,
      "scope": "首轮对应自动化断言；引用PN按引用条目计，非任务数",
      "status": "MEASURED"
    },
    {
      "id": "fault_tolerance",
      "mode": "offline",
      "name": "故障后回答可用",
      "value": 0.75,
      "numerator": 24,
      "denominator": 32,
      "unit": "ratio",
      "target": 0.85,
      "scope": "4类故障×8类Persona首轮固定探针",
      "status": "MEASURED"
    },
    {
      "id": "degraded_reason",
      "mode": "offline",
      "name": "显式降级原因",
      "value": 0.0,
      "numerator": 0,
      "denominator": 32,
      "unit": "ratio",
      "target": 1,
      "scope": "4类故障×8类Persona首轮固定探针",
      "status": "MEASURED"
    },
    {
      "id": "faithfulness",
      "mode": "offline",
      "name": "支持主张比例（模型估计）",
      "value": null,
      "numerator": 0,
      "denominator": 0,
      "unit": "ratio",
      "target": 0.95,
      "scope": "首轮独立提示词，同模型裁判；非官方Ragas运行",
      "status": "NOT_MEASURED"
    },
    {
      "id": "unsupported_answer",
      "mode": "offline",
      "name": "含不支持主张的轨迹比例（模型估计）",
      "value": null,
      "numerator": 0,
      "denominator": 0,
      "unit": "ratio",
      "target": 0.02,
      "scope": "按完整轨迹计，不等同于人工确认的编造事实率",
      "status": "NOT_MEASURED"
    },
    {
      "id": "goal",
      "mode": "offline",
      "name": "语义目标达成（模型估计）",
      "value": null,
      "numerator": 0,
      "denominator": 0,
      "unit": "ratio",
      "target": null,
      "scope": "首轮可评回复；排除无回复与裁判错误，不计为通过",
      "status": "NOT_MEASURED"
    },
    {
      "id": "joint",
      "mode": "offline",
      "name": "硬断言+裁判联合通过",
      "value": null,
      "numerator": 0,
      "denominator": 0,
      "unit": "ratio",
      "target": null,
      "scope": "仅有有效裁判结果的首轮轨迹，不用于pass^8",
      "status": "NOT_MEASURED"
    },
    {
      "id": "tone",
      "mode": "offline",
      "name": "tone",
      "value": null,
      "numerator": null,
      "denominator": 0,
      "unit": "score_1_5",
      "target": 4,
      "scope": "首轮模型裁判1–5分；故障兜底仅取注入场景",
      "status": "NOT_MEASURED"
    },
    {
      "id": "smoothness",
      "mode": "offline",
      "name": "smoothness",
      "value": null,
      "numerator": null,
      "denominator": 0,
      "unit": "score_1_5",
      "target": 4,
      "scope": "首轮模型裁判1–5分；故障兜底仅取注入场景",
      "status": "NOT_MEASURED"
    },
    {
      "id": "clarity",
      "mode": "offline",
      "name": "clarity",
      "value": null,
      "numerator": null,
      "denominator": 0,
      "unit": "score_1_5",
      "target": 4,
      "scope": "首轮模型裁判1–5分；故障兜底仅取注入场景",
      "status": "NOT_MEASURED"
    },
    {
      "id": "relevance",
      "mode": "offline",
      "name": "relevance",
      "value": null,
      "numerator": null,
      "denominator": 0,
      "unit": "score_1_5",
      "target": 4,
      "scope": "首轮模型裁判1–5分；故障兜底仅取注入场景",
      "status": "NOT_MEASURED"
    },
    {
      "id": "fallback_quality",
      "mode": "offline",
      "name": "fallback_quality",
      "value": null,
      "numerator": null,
      "denominator": 0,
      "unit": "score_1_5",
      "target": 4,
      "scope": "首轮模型裁判1–5分；故障兜底仅取注入场景",
      "status": "NOT_MEASURED"
    },
    {
      "id": "latency_p50",
      "mode": "offline",
      "name": "答案时延P50",
      "value": 2.66,
      "unit": "ms",
      "denominator": 736,
      "target": 3000,
      "status": "MEASURED",
      "scope": "无人工故障，应用处理+真实模型API（线上）；不含浏览器渲染。nearest-rank分位数。"
    },
    {
      "id": "latency_p95",
      "mode": "offline",
      "name": "答案时延P95",
      "value": 11.53,
      "unit": "ms",
      "denominator": 736,
      "target": 8000,
      "status": "MEASURED",
      "scope": "无人工故障，应用处理+真实模型API（线上）；不含浏览器渲染。nearest-rank分位数。"
    },
    {
      "id": "fcr",
      "mode": "offline",
      "name": "7日一次解决率",
      "value": null,
      "status": "NOT_MEASURED",
      "target": 0.7,
      "reason": "没有7日联系与重开观测数据；不能用短对话成功率替代"
    },
    {
      "id": "hard_pass",
      "mode": "online",
      "name": "核心硬断言通过率",
      "value": 0.8378378378378378,
      "numerator": 62,
      "denominator": 74,
      "unit": "ratio",
      "target": null,
      "scope": "74个核心任务首轮",
      "status": "MEASURED"
    },
    {
      "id": "pass_1",
      "mode": "online",
      "name": "pass^1",
      "value": 0.8378378378378378,
      "numerator": 62,
      "denominator": 74,
      "target": null,
      "unit": "ratio",
      "status": "MEASURED",
      "scope": "核心任务前k次全部硬断言通过；不含重复语义质量"
    },
    {
      "id": "pass_3",
      "mode": "online",
      "name": "pass^3",
      "value": 0.8378378378378378,
      "numerator": 62,
      "denominator": 74,
      "target": 0.85,
      "unit": "ratio",
      "status": "MEASURED",
      "scope": "核心任务前k次全部硬断言通过；不含重复语义质量"
    },
    {
      "id": "pass_8",
      "mode": "online",
      "name": "pass^8",
      "value": 0.8378378378378378,
      "numerator": 62,
      "denominator": 74,
      "target": 0.6,
      "unit": "ratio",
      "status": "MEASURED",
      "scope": "核心任务前k次全部硬断言通过；不含重复语义质量"
    },
    {
      "id": "order",
      "mode": "online",
      "name": "订单字段完整且一致",
      "value": 1.0,
      "numerator": 12,
      "denominator": 12,
      "unit": "ratio",
      "target": 1,
      "scope": "首轮对应自动化断言；引用PN按引用条目计，非任务数",
      "status": "MEASURED"
    },
    {
      "id": "pn",
      "mode": "online",
      "name": "引用PN一致",
      "value": 1.0,
      "numerator": 26,
      "denominator": 26,
      "unit": "ratio",
      "target": 1,
      "scope": "首轮对应自动化断言；引用PN按引用条目计，非任务数",
      "status": "MEASURED"
    },
    {
      "id": "auth",
      "mode": "online",
      "name": "动作边界遵循",
      "value": 1.0,
      "numerator": 14,
      "denominator": 14,
      "unit": "ratio",
      "target": 1,
      "scope": "首轮对应自动化断言；引用PN按引用条目计，非任务数",
      "status": "MEASURED"
    },
    {
      "id": "robust",
      "mode": "online",
      "name": "等义扰动终态一致",
      "value": 1.0,
      "numerator": 3,
      "denominator": 3,
      "unit": "ratio",
      "target": 0.8,
      "scope": "首轮对应自动化断言；引用PN按引用条目计，非任务数",
      "status": "MEASURED"
    },
    {
      "id": "handoff",
      "mode": "online",
      "name": "转人工上下文字段完整",
      "value": 0.0,
      "numerator": 0,
      "denominator": 1,
      "unit": "ratio",
      "target": 1,
      "scope": "首轮对应自动化断言；引用PN按引用条目计，非任务数",
      "status": "MEASURED"
    },
    {
      "id": "restore",
      "mode": "online",
      "name": "同进程会话恢复",
      "value": 1.0,
      "numerator": 1,
      "denominator": 1,
      "unit": "ratio",
      "target": 0.9,
      "scope": "首轮对应自动化断言；引用PN按引用条目计，非任务数",
      "status": "MEASURED"
    },
    {
      "id": "fault_tolerance",
      "mode": "online",
      "name": "故障后回答可用",
      "value": 0.75,
      "numerator": 24,
      "denominator": 32,
      "unit": "ratio",
      "target": 0.85,
      "scope": "4类故障×8类Persona首轮固定探针",
      "status": "MEASURED"
    },
    {
      "id": "degraded_reason",
      "mode": "online",
      "name": "显式降级原因",
      "value": 0.0,
      "numerator": 0,
      "denominator": 32,
      "unit": "ratio",
      "target": 1,
      "scope": "4类故障×8类Persona首轮固定探针",
      "status": "MEASURED"
    },
    {
      "id": "faithfulness",
      "mode": "online",
      "name": "支持主张比例（模型估计）",
      "value": 0.8603988603988604,
      "numerator": 302,
      "denominator": 351,
      "unit": "ratio",
      "target": 0.95,
      "scope": "首轮独立提示词，同模型裁判；非官方Ragas运行",
      "status": "MEASURED"
    },
    {
      "id": "unsupported_answer",
      "mode": "online",
      "name": "含不支持主张的轨迹比例（模型估计）",
      "value": 0.40594059405940597,
      "numerator": 41,
      "denominator": 101,
      "unit": "ratio",
      "target": 0.02,
      "scope": "按完整轨迹计，不等同于人工确认的编造事实率",
      "status": "MEASURED"
    },
    {
      "id": "goal",
      "mode": "online",
      "name": "语义目标达成（模型估计）",
      "value": 0.7326732673267327,
      "numerator": 74,
      "denominator": 101,
      "unit": "ratio",
      "target": null,
      "scope": "首轮可评回复；排除无回复与裁判错误，不计为通过",
      "status": "MEASURED"
    },
    {
      "id": "joint",
      "mode": "online",
      "name": "硬断言+裁判联合通过",
      "value": 0.6831683168316832,
      "numerator": 69,
      "denominator": 101,
      "unit": "ratio",
      "target": null,
      "scope": "仅有有效裁判结果的首轮轨迹，不用于pass^8",
      "status": "MEASURED"
    },
    {
      "id": "tone",
      "mode": "online",
      "name": "tone",
      "value": 4.297029702970297,
      "numerator": null,
      "denominator": 101,
      "unit": "score_1_5",
      "target": 4,
      "scope": "首轮模型裁判1–5分；故障兜底仅取注入场景",
      "status": "MEASURED"
    },
    {
      "id": "smoothness",
      "mode": "online",
      "name": "smoothness",
      "value": 4.297029702970297,
      "numerator": null,
      "denominator": 101,
      "unit": "score_1_5",
      "target": 4,
      "scope": "首轮模型裁判1–5分；故障兜底仅取注入场景",
      "status": "MEASURED"
    },
    {
      "id": "clarity",
      "mode": "online",
      "name": "clarity",
      "value": 4.415841584158416,
      "numerator": null,
      "denominator": 101,
      "unit": "score_1_5",
      "target": 4,
      "scope": "首轮模型裁判1–5分；故障兜底仅取注入场景",
      "status": "MEASURED"
    },
    {
      "id": "relevance",
      "mode": "online",
      "name": "relevance",
      "value": 4.227722772277228,
      "numerator": null,
      "denominator": 101,
      "unit": "score_1_5",
      "target": 4,
      "scope": "首轮模型裁判1–5分；故障兜底仅取注入场景",
      "status": "MEASURED"
    },
    {
      "id": "fallback_quality",
      "mode": "online",
      "name": "fallback_quality",
      "value": 2.3333333333333335,
      "numerator": null,
      "denominator": 3,
      "unit": "score_1_5",
      "target": 4,
      "scope": "首轮模型裁判1–5分；故障兜底仅取注入场景",
      "status": "MEASURED"
    },
    {
      "id": "latency_p50",
      "mode": "online",
      "name": "答案时延P50",
      "value": 2629.68,
      "unit": "ms",
      "denominator": 746,
      "target": 3000,
      "status": "MEASURED",
      "scope": "无人工故障，应用处理+真实模型API（线上）；不含浏览器渲染。nearest-rank分位数。"
    },
    {
      "id": "latency_p95",
      "mode": "online",
      "name": "答案时延P95",
      "value": 8999.15,
      "unit": "ms",
      "denominator": 746,
      "target": 8000,
      "status": "MEASURED",
      "scope": "无人工故障，应用处理+真实模型API（线上）；不含浏览器渲染。nearest-rank分位数。"
    },
    {
      "id": "fcr",
      "mode": "online",
      "name": "7日一次解决率",
      "value": null,
      "status": "NOT_MEASURED",
      "target": 0.7,
      "reason": "没有7日联系与重开观测数据；不能用短对话成功率替代"
    }
  ],
  "target_note": "全部目标来自本项目验收方法，不是行业公认统一标准",
  "semantic_limits": "裁判同模型且未做人工校准；仅首轮；上下文中含真实规则以免把正确拒绝误判成无依据。"
}