{
  "schema": "lab.taskDiscrimination/1",
  "generatedAt": "2026-08-26T02:07:32.960Z",
  "step": "区分度实验（第 11 步 §8 第 6 条的销账）",
  "baselineKey": {
    "subjectSha": "4a68e02d3d695cee3c39d255ea5d20f7f2bf9b8c",
    "modelId": "claude-opus-5",
    "promptTemplateVersion": "pt-v1"
  },
  "tasks": [
    "GT-001",
    "GT-002",
    "GT-003"
  ],
  "groupSizes": {
    "GT-001": 5,
    "GT-002": 5,
    "GT-003": 5
  },
  "minN": 5,
  "allGroupsEligible": true,
  "totalRuns": 15,
  "promptHashByTask": {
    "GT-001": "559c6decc5b4cf58d12e2062aed55810d67367ecca16a9761b019c4275813d14",
    "GT-002": "7d1255217ed1bdb90bca3b0181d61afb119f3fc663edef979748f3d18ae6da28",
    "GT-003": "f96b13a7cc7a2da327a494e2bb33b047915113c61b8343bb227d31c29e212a50"
  },
  "noiseFloorSources": {
    "GT-001": "/Users/nancy/Documents/AICoding/VS_Code/ai-ds-lab/subjects/tvu-ds/inventory/noise-floor-gt001-n5-4a68e02d.json",
    "GT-002": "/Users/nancy/Documents/AICoding/VS_Code/ai-ds-lab/subjects/tvu-ds/inventory/noise-floor-gt002-n5-4a68e02d.json",
    "GT-003": "/Users/nancy/Documents/AICoding/VS_Code/ai-ds-lab/subjects/tvu-ds/inventory/noise-floor-gt003-n5-4a68e02d.json"
  },
  "ruleSurface": {
    "gauge": "audit-mockup-html-conformance",
    "K": 5,
    "byId": {
      "popupbox-for-confirm": "error",
      "unconstrained-svg": "error",
      "handrolled-nav": "error",
      "hardcoded-color": "warn",
      "alert-confirmText": "warn"
    }
  },
  "perRun": [
    {
      "runId": "r001",
      "taskId": "GT-001",
      "repeatIndex": 1,
      "exitCode": 1,
      "errors": 1,
      "warnings": 0,
      "violatedRules": [
        "popupbox-for-confirm"
      ],
      "htmlBytes": 74834,
      "escapeRate": 0.3759,
      "unmatchedDistinct": 18
    },
    {
      "runId": "r002",
      "taskId": "GT-001",
      "repeatIndex": 2,
      "exitCode": 1,
      "errors": 1,
      "warnings": 0,
      "violatedRules": [
        "popupbox-for-confirm"
      ],
      "htmlBytes": 85315,
      "escapeRate": 0.408,
      "unmatchedDistinct": 19
    },
    {
      "runId": "r003",
      "taskId": "GT-001",
      "repeatIndex": 3,
      "exitCode": 1,
      "errors": 4,
      "warnings": 0,
      "violatedRules": [
        "popupbox-for-confirm"
      ],
      "htmlBytes": 78343,
      "escapeRate": 0.4115,
      "unmatchedDistinct": 17
    },
    {
      "runId": "r004",
      "taskId": "GT-001",
      "repeatIndex": 4,
      "exitCode": 0,
      "errors": 0,
      "warnings": 0,
      "violatedRules": [],
      "htmlBytes": 73095,
      "escapeRate": 0.3425,
      "unmatchedDistinct": 21
    },
    {
      "runId": "r005",
      "taskId": "GT-001",
      "repeatIndex": 5,
      "exitCode": 1,
      "errors": 2,
      "warnings": 0,
      "violatedRules": [
        "popupbox-for-confirm"
      ],
      "htmlBytes": 77187,
      "escapeRate": 0.4171,
      "unmatchedDistinct": 16
    },
    {
      "runId": "r007",
      "taskId": "GT-002",
      "repeatIndex": 1,
      "exitCode": 1,
      "errors": 1,
      "warnings": 0,
      "violatedRules": [
        "handrolled-nav"
      ],
      "htmlBytes": 40900,
      "escapeRate": 0.3708,
      "unmatchedDistinct": 15
    },
    {
      "runId": "r008",
      "taskId": "GT-002",
      "repeatIndex": 2,
      "exitCode": 1,
      "errors": 1,
      "warnings": 0,
      "violatedRules": [
        "handrolled-nav"
      ],
      "htmlBytes": 34129,
      "escapeRate": 0.3054,
      "unmatchedDistinct": 7
    },
    {
      "runId": "r009",
      "taskId": "GT-002",
      "repeatIndex": 3,
      "exitCode": 1,
      "errors": 4,
      "warnings": 0,
      "violatedRules": [
        "handrolled-nav"
      ],
      "htmlBytes": 50677,
      "escapeRate": 0.3149,
      "unmatchedDistinct": 4
    },
    {
      "runId": "r010",
      "taskId": "GT-002",
      "repeatIndex": 4,
      "exitCode": 0,
      "errors": 0,
      "warnings": 0,
      "violatedRules": [],
      "htmlBytes": 80078,
      "escapeRate": 0.3293,
      "unmatchedDistinct": 18
    },
    {
      "runId": "r011",
      "taskId": "GT-002",
      "repeatIndex": 5,
      "exitCode": 1,
      "errors": 4,
      "warnings": 0,
      "violatedRules": [
        "handrolled-nav"
      ],
      "htmlBytes": 51203,
      "escapeRate": 0.3054,
      "unmatchedDistinct": 3
    },
    {
      "runId": "r012",
      "taskId": "GT-003",
      "repeatIndex": 1,
      "exitCode": 0,
      "errors": 0,
      "warnings": 0,
      "violatedRules": [],
      "htmlBytes": 77034,
      "escapeRate": 0.4104,
      "unmatchedDistinct": 12
    },
    {
      "runId": "r013",
      "taskId": "GT-003",
      "repeatIndex": 2,
      "exitCode": 0,
      "errors": 0,
      "warnings": 0,
      "violatedRules": [],
      "htmlBytes": 70152,
      "escapeRate": 0.5172,
      "unmatchedDistinct": 11
    },
    {
      "runId": "r014",
      "taskId": "GT-003",
      "repeatIndex": 3,
      "exitCode": 1,
      "errors": 1,
      "warnings": 0,
      "violatedRules": [
        "unconstrained-svg"
      ],
      "htmlBytes": 64294,
      "escapeRate": 0.3485,
      "unmatchedDistinct": 10
    },
    {
      "runId": "r015",
      "taskId": "GT-003",
      "repeatIndex": 4,
      "exitCode": 1,
      "errors": 1,
      "warnings": 0,
      "violatedRules": [
        "handrolled-nav"
      ],
      "htmlBytes": 75931,
      "escapeRate": 0.3974,
      "unmatchedDistinct": 15
    },
    {
      "runId": "r017",
      "taskId": "GT-003",
      "repeatIndex": 5,
      "exitCode": 1,
      "errors": 1,
      "warnings": 0,
      "violatedRules": [
        "handrolled-nav"
      ],
      "htmlBytes": 45438,
      "escapeRate": 0.3771,
      "unmatchedDistinct": 20
    }
  ],
  "perTaskStats": {
    "GT-001": {
      "ds-conformance.errors（严口径：只数 error）": {
        "n": 5,
        "mean": 1.6,
        "sd": 1.5166,
        "min": 0,
        "max": 4,
        "values": [
          1,
          1,
          4,
          0,
          2
        ]
      },
      "ds-conformance.exitCode（run 级二值 pass/fail）": {
        "n": 5,
        "mean": 0.2,
        "sd": 0.4472,
        "min": 0,
        "max": 1,
        "values": [
          0,
          0,
          0,
          1,
          0
        ]
      },
      "escape-rate.escapeRate（比率口径）": {
        "n": 5,
        "mean": 0.391,
        "sd": 0.0315,
        "min": 0.3425,
        "max": 0.4171,
        "values": [
          0.3759,
          0.408,
          0.4115,
          0.3425,
          0.4171
        ]
      },
      "escape-rate.unmatchedDistinct（DS 缺的洞口径）": {
        "n": 5,
        "mean": 18.2,
        "sd": 1.9235,
        "min": 16,
        "max": 21,
        "values": [
          18,
          19,
          17,
          21,
          16
        ]
      },
      "output.htmlBytes": {
        "n": 5,
        "mean": 77754.8,
        "sd": 4692.2758,
        "min": 73095,
        "max": 85315,
        "values": [
          74834,
          85315,
          78343,
          73095,
          77187
        ]
      }
    },
    "GT-002": {
      "ds-conformance.errors（严口径：只数 error）": {
        "n": 5,
        "mean": 2,
        "sd": 1.8708,
        "min": 0,
        "max": 4,
        "values": [
          1,
          1,
          4,
          0,
          4
        ]
      },
      "ds-conformance.exitCode（run 级二值 pass/fail）": {
        "n": 5,
        "mean": 0.2,
        "sd": 0.4472,
        "min": 0,
        "max": 1,
        "values": [
          0,
          0,
          0,
          1,
          0
        ]
      },
      "escape-rate.escapeRate（比率口径）": {
        "n": 5,
        "mean": 0.3252,
        "sd": 0.0273,
        "min": 0.3054,
        "max": 0.3708,
        "values": [
          0.3708,
          0.3054,
          0.3149,
          0.3293,
          0.3054
        ]
      },
      "escape-rate.unmatchedDistinct（DS 缺的洞口径）": {
        "n": 5,
        "mean": 9.4,
        "sd": 6.7305,
        "min": 3,
        "max": 18,
        "values": [
          15,
          7,
          4,
          18,
          3
        ]
      },
      "output.htmlBytes": {
        "n": 5,
        "mean": 51397.4,
        "sd": 17546.5569,
        "min": 34129,
        "max": 80078,
        "values": [
          40900,
          34129,
          50677,
          80078,
          51203
        ]
      }
    },
    "GT-003": {
      "ds-conformance.errors（严口径：只数 error）": {
        "n": 5,
        "mean": 0.6,
        "sd": 0.5477,
        "min": 0,
        "max": 1,
        "values": [
          0,
          0,
          1,
          1,
          1
        ]
      },
      "ds-conformance.exitCode（run 级二值 pass/fail）": {
        "n": 5,
        "mean": 0.4,
        "sd": 0.5477,
        "min": 0,
        "max": 1,
        "values": [
          1,
          1,
          0,
          0,
          0
        ]
      },
      "escape-rate.escapeRate（比率口径）": {
        "n": 5,
        "mean": 0.4101,
        "sd": 0.0643,
        "min": 0.3485,
        "max": 0.5172,
        "values": [
          0.4104,
          0.5172,
          0.3485,
          0.3974,
          0.3771
        ]
      },
      "escape-rate.unmatchedDistinct（DS 缺的洞口径）": {
        "n": 5,
        "mean": 13.6,
        "sd": 4.0373,
        "min": 10,
        "max": 20,
        "values": [
          12,
          11,
          10,
          15,
          20
        ]
      },
      "output.htmlBytes": {
        "n": 5,
        "mean": 66569.8,
        "sd": 12861.7256,
        "min": 45438,
        "max": 77034,
        "values": [
          77034,
          70152,
          64294,
          75931,
          45438
        ]
      }
    }
  },
  "discrimination": {
    "ds-conformance.errors（严口径：只数 error）": {
      "type": "count",
      "pairs": {
        "GT-001 vs GT-002": {
          "meanA": 1.6,
          "meanB": 2,
          "deltaMean": 0.4,
          "primaryTest": "Mann-Whitney U 精确检验（枚举秩分配）",
          "p": 0.9762,
          "testDetail": {
            "W": 26.5,
            "EW": 27.5,
            "p": 0.9762,
            "exact": true,
            "enumerated": 252,
            "tiedObservations": 9,
            "note": null
          },
          "discriminable": false,
          "mdePooled": 3.0174,
          "mdeRatio": 0.1326,
          "note": "⚠️ discriminable 以主检验 p 为准；mdeRatio 只作并排读数，⛔ 不得当显著性用（边界 ⑤）"
        },
        "GT-001 vs GT-003": {
          "meanA": 1.6,
          "meanB": 0.6,
          "deltaMean": 1,
          "primaryTest": "Mann-Whitney U 精确检验（枚举秩分配）",
          "p": 0.3968,
          "testDetail": {
            "W": 33,
            "EW": 27.5,
            "p": 0.3968,
            "exact": true,
            "enumerated": 252,
            "tiedObservations": 8,
            "note": null
          },
          "discriminable": false,
          "mdePooled": 2.0203,
          "mdeRatio": 0.495,
          "note": "⚠️ discriminable 以主检验 p 为准；mdeRatio 只作并排读数，⛔ 不得当显著性用（边界 ⑤）"
        },
        "GT-002 vs GT-003": {
          "meanA": 2,
          "meanB": 0.6,
          "deltaMean": 1.4,
          "primaryTest": "Mann-Whitney U 精确检验（枚举秩分配）",
          "p": 0.3968,
          "testDetail": {
            "W": 33,
            "EW": 27.5,
            "p": 0.3968,
            "exact": true,
            "enumerated": 252,
            "tiedObservations": 10,
            "note": null
          },
          "discriminable": false,
          "mdePooled": 2.4423,
          "mdeRatio": 0.5732,
          "note": "⚠️ discriminable 以主检验 p 为准；mdeRatio 只作并排读数，⛔ 不得当显著性用（边界 ⑤）"
        }
      }
    },
    "ds-conformance.exitCode（run 级二值 pass/fail）": {
      "type": "binary",
      "pairs": {
        "GT-001 vs GT-002": {
          "meanA": 0.2,
          "meanB": 0.2,
          "deltaMean": 0,
          "primaryTest": "Fisher 精确检验（α=0.05 双侧）",
          "p": 1,
          "testDetail": {
            "passA": 1,
            "nA": 5,
            "passB": 1,
            "nB": 5
          },
          "discriminable": false,
          "mdePooled": null,
          "mdeRatio": null,
          "note": "⚠️ discriminable 以主检验 p 为准；mdeRatio 只作并排读数，⛔ 不得当显著性用（边界 ⑤）"
        },
        "GT-001 vs GT-003": {
          "meanA": 0.2,
          "meanB": 0.4,
          "deltaMean": 0.2,
          "primaryTest": "Fisher 精确检验（α=0.05 双侧）",
          "p": 1,
          "testDetail": {
            "passA": 1,
            "nA": 5,
            "passB": 2,
            "nB": 5
          },
          "discriminable": false,
          "mdePooled": null,
          "mdeRatio": null,
          "note": "⚠️ discriminable 以主检验 p 为准；mdeRatio 只作并排读数，⛔ 不得当显著性用（边界 ⑤）"
        },
        "GT-002 vs GT-003": {
          "meanA": 0.2,
          "meanB": 0.4,
          "deltaMean": 0.2,
          "primaryTest": "Fisher 精确检验（α=0.05 双侧）",
          "p": 1,
          "testDetail": {
            "passA": 1,
            "nA": 5,
            "passB": 2,
            "nB": 5
          },
          "discriminable": false,
          "mdePooled": null,
          "mdeRatio": null,
          "note": "⚠️ discriminable 以主检验 p 为准；mdeRatio 只作并排读数，⛔ 不得当显著性用（边界 ⑤）"
        }
      }
    },
    "escape-rate.escapeRate（比率口径）": {
      "type": "continuous",
      "pairs": {
        "GT-001 vs GT-002": {
          "meanA": 0.391,
          "meanB": 0.3252,
          "deltaMean": 0.0658,
          "primaryTest": "Welch t 检验（不假设等方差）",
          "p": 0.008,
          "testDetail": {
            "t": 3.5301,
            "df": 7.843,
            "p": 0.008,
            "note": null
          },
          "discriminable": true,
          "mdePooled": 0.0522,
          "mdeRatio": 1.2605,
          "note": "⚠️ discriminable 以主检验 p 为准；mdeRatio 只作并排读数，⛔ 不得当显著性用（边界 ⑤）"
        },
        "GT-001 vs GT-003": {
          "meanA": 0.391,
          "meanB": 0.4101,
          "deltaMean": 0.0191,
          "primaryTest": "Welch t 检验（不假设等方差）",
          "p": 0.5727,
          "testDetail": {
            "t": -0.5974,
            "df": 5.8183,
            "p": 0.5727,
            "note": null
          },
          "discriminable": false,
          "mdePooled": 0.0897,
          "mdeRatio": 0.2129,
          "note": "⚠️ discriminable 以主检验 p 为准；mdeRatio 只作并排读数，⛔ 不得当显著性用（边界 ⑤）"
        },
        "GT-002 vs GT-003": {
          "meanA": 0.3252,
          "meanB": 0.4101,
          "deltaMean": 0.0849,
          "primaryTest": "Welch t 检验（不假设等方差）",
          "p": 0.0385,
          "testDetail": {
            "t": -2.7208,
            "df": 5.4008,
            "p": 0.0385,
            "note": null
          },
          "discriminable": true,
          "mdePooled": 0.0875,
          "mdeRatio": 0.9703,
          "note": "⚠️ discriminable 以主检验 p 为准；mdeRatio 只作并排读数，⛔ 不得当显著性用（边界 ⑤）"
        }
      }
    },
    "escape-rate.unmatchedDistinct（DS 缺的洞口径）": {
      "type": "count",
      "pairs": {
        "GT-001 vs GT-002": {
          "meanA": 18.2,
          "meanB": 9.4,
          "deltaMean": 8.8,
          "primaryTest": "Mann-Whitney U 精确检验（枚举秩分配）",
          "p": 0.0397,
          "testDetail": {
            "W": 37.5,
            "EW": 27.5,
            "p": 0.0397,
            "exact": true,
            "enumerated": 252,
            "tiedObservations": 2,
            "note": null
          },
          "discriminable": true,
          "mdePooled": 8.7703,
          "mdeRatio": 1.0034,
          "note": "⚠️ discriminable 以主检验 p 为准；mdeRatio 只作并排读数，⛔ 不得当显著性用（边界 ⑤）"
        },
        "GT-001 vs GT-003": {
          "meanA": 18.2,
          "meanB": 13.6,
          "deltaMean": 4.6,
          "primaryTest": "Mann-Whitney U 精确检验（枚举秩分配）",
          "p": 0.0952,
          "testDetail": {
            "W": 36,
            "EW": 27.5,
            "p": 0.0952,
            "exact": true,
            "enumerated": 252,
            "tiedObservations": 0,
            "note": null
          },
          "discriminable": false,
          "mdePooled": 5.6031,
          "mdeRatio": 0.821,
          "note": "⚠️ discriminable 以主检验 p 为准；mdeRatio 只作并排读数，⛔ 不得当显著性用（边界 ⑤）"
        },
        "GT-002 vs GT-003": {
          "meanA": 9.4,
          "meanB": 13.6,
          "deltaMean": 4.2,
          "primaryTest": "Mann-Whitney U 精确检验（枚举秩分配）",
          "p": 0.3333,
          "testDetail": {
            "W": 22.5,
            "EW": 27.5,
            "p": 0.3333,
            "exact": true,
            "enumerated": 252,
            "tiedObservations": 2,
            "note": null
          },
          "discriminable": false,
          "mdePooled": 9.8335,
          "mdeRatio": 0.4271,
          "note": "⚠️ discriminable 以主检验 p 为准；mdeRatio 只作并排读数，⛔ 不得当显著性用（边界 ⑤）"
        }
      }
    },
    "output.htmlBytes": {
      "type": "continuous",
      "pairs": {
        "GT-001 vs GT-002": {
          "meanA": 77754.8,
          "meanB": 51397.4,
          "deltaMean": 26357.4,
          "primaryTest": "Welch t 检验（不假设等方差）",
          "p": 0.026,
          "testDetail": {
            "t": 3.2449,
            "df": 4.5692,
            "p": 0.026,
            "note": null
          },
          "discriminable": true,
          "mdePooled": 22756.7047,
          "mdeRatio": 1.1582,
          "note": "⚠️ discriminable 以主检验 p 为准；mdeRatio 只作并排读数，⛔ 不得当显著性用（边界 ⑤）"
        },
        "GT-001 vs GT-003": {
          "meanA": 77754.8,
          "meanB": 66569.8,
          "deltaMean": 11185,
          "primaryTest": "Welch t 检验（不假设等方差）",
          "p": 0.1268,
          "testDetail": {
            "t": 1.8268,
            "df": 5.0462,
            "p": 0.1268,
            "note": null
          },
          "discriminable": false,
          "mdePooled": 17153.4549,
          "mdeRatio": 0.6521,
          "note": "⚠️ discriminable 以主检验 p 为准；mdeRatio 只作并排读数，⛔ 不得当显著性用（边界 ⑤）"
        },
        "GT-002 vs GT-003": {
          "meanA": 51397.4,
          "meanB": 66569.8,
          "deltaMean": 15172.4,
          "primaryTest": "Welch t 检验（不假设等方差）",
          "p": 0.1609,
          "testDetail": {
            "t": -1.5594,
            "df": 7.3355,
            "p": 0.1609,
            "note": null
          },
          "discriminable": false,
          "mdePooled": 27257.7274,
          "mdeRatio": 0.5566,
          "note": "⚠️ discriminable 以主检验 p 为准；mdeRatio 只作并排读数，⛔ 不得当显著性用（边界 ⑤）"
        }
      }
    }
  },
  "exposureMatrix": {
    "matrix": {
      "GT-001": {
        "alert-confirmText": 0,
        "handrolled-nav": 0,
        "hardcoded-color": 0,
        "popupbox-for-confirm": 4,
        "unconstrained-svg": 0
      },
      "GT-002": {
        "alert-confirmText": 0,
        "handrolled-nav": 4,
        "hardcoded-color": 0,
        "popupbox-for-confirm": 0,
        "unconstrained-svg": 0
      },
      "GT-003": {
        "alert-confirmText": 0,
        "handrolled-nav": 2,
        "hardcoded-color": 0,
        "popupbox-for-confirm": 0,
        "unconstrained-svg": 1
      }
    },
    "everExposed": [
      "handrolled-nav",
      "popupbox-for-confirm",
      "unconstrained-svg"
    ],
    "neverExposed": [
      "alert-confirmText",
      "hardcoded-color"
    ],
    "effectiveDenominator": {
      "name": "曾曝光规则数（15 run · 3 archetype 口径）",
      "value": 3,
      "of": 5
    },
    "boundary": "⛔ `neverExposed` 不得读作「死规则」—— 只能读作「本 set、本 N 下未曾曝光」（AGENTS §2.7 / 边界 ④）"
  },
  "preRegistered": [
    {
      "id": "P1",
      "task": "GT-002",
      "claim": "GT-002 上 `popupbox-for-confirm` 的命中 run 数 = 0",
      "formalizedBy": "lab",
      "formalizationNote": "⚠️ `claim` 是 lab 对 `sourceQuote` 那句自然语言的**形式化**，⛔ 不是 task 文件的原话",
      "sourceQuote": "表单面主要打 unconstrained-svg 与 hardcoded-color；不含确认弹窗 ⇒ **popupbox-for-confirm 在本条上应恒不命中**，可作为跨任务的 must-not-hit 对照。",
      "quoteMustContain": "popupbox-for-confirm 在本条上应恒不命中",
      "quoteVerified": true,
      "result": {
        "actual": 0,
        "hit": true
      }
    },
    {
      "id": "P2",
      "task": "GT-002",
      "claim": "GT-002 上 `unconstrained-svg` 或 `hardcoded-color` 的命中 run 数 > 0",
      "formalizedBy": "lab",
      "formalizationNote": "⚠️ `claim` 是 lab 对 `sourceQuote` 那句自然语言的**形式化**，⛔ 不是 task 文件的原话",
      "sourceQuote": "表单面主要打 unconstrained-svg 与 hardcoded-color；不含确认弹窗 ⇒ **popupbox-for-confirm 在本条上应恒不命中**，可作为跨任务的 must-not-hit 对照。",
      "quoteMustContain": "表单面主要打 unconstrained-svg 与 hardcoded-color",
      "quoteVerified": true,
      "result": {
        "actual": 0,
        "hit": false
      }
    },
    {
      "id": "P3",
      "task": "GT-003",
      "claim": "GT-003 上 (`unconstrained-svg` + `hardcoded-color`) 的命中 run 数 ≥ GT-001 与 GT-002 各自的同一量",
      "formalizedBy": "lab",
      "formalizationNote": "⚠️ `claim` 是 lab 对 `sourceQuote` 那句自然语言的**形式化**，⛔ 不是 task 文件的原话",
      "sourceQuote": "图表要么自画 inline <svg>（打 unconstrained-svg），要么自定色（打 hardcoded-color）⇒ 这条任务对这两条规则的曝光最强。",
      "quoteMustContain": "这条任务对这两条规则的曝光最强",
      "quoteVerified": true,
      "result": {
        "actual": 1,
        "others": [
          {
            "task": "GT-001",
            "v": 0
          },
          {
            "task": "GT-002",
            "v": 0
          }
        ],
        "hit": true
      }
    },
    {
      "id": "P4",
      "task": "GT-003",
      "claim": "GT-003 的 `escape-rate.unmatchedDistinct` 均值 > GT-001 与 GT-002 各自的均值",
      "formalizedBy": "lab",
      "formalizationNote": "⚠️ `claim` 是 lab 对 `sourceQuote` 那句自然语言的**形式化**，⛔ 不是 task 文件的原话",
      "sourceQuote": "**本条是三条里 escape-rate 的主力样本**：DS 自陈 chart primitives 缺位 ⇒ 图表配色只能逃逸，逃逸值里 `sameValueToken` 为空的那类（= 真源缺的洞）分子最足。",
      "quoteMustContain": "本条是三条里 escape-rate 的主力样本",
      "quoteVerified": true,
      "result": {
        "actual": 13.6,
        "others": [
          {
            "task": "GT-001",
            "v": 18.2
          },
          {
            "task": "GT-002",
            "v": 9.4
          }
        ],
        "hit": false
      }
    }
  ],
  "controls": [
    {
      "id": "MH1",
      "ok": true,
      "detail": "规则面必须恰好 5 条且 id 集合逐字对上（钉具名事实，⛔ 不钉「解析器跑通」）—— 实测 K=5：alert-confirmText,handrolled-nav,hardcoded-color,popupbox-for-confirm,unconstrained-svg"
    },
    {
      "id": "MH2",
      "ok": true,
      "detail": "必须恰好 3 个 taskId 且每个恰好 5 个 run —— 实测 3 个：{\"GT-001\":5,\"GT-002\":5,\"GT-003\":5}"
    },
    {
      "id": "MH3",
      "ok": true,
      "detail": "每个 task 的 promptHash 唯一，且三个互不相同（否则不是三条不同 task）—— 实测 GT-001=559c6dec GT-002=7d125521 GT-003=f96b13a7"
    },
    {
      "id": "MH4",
      "ok": true,
      "detail": "§5.2 三元组必须在全部 15 个 run 上唯一且非空（否则区分度里混进了别的变量）—— 实测 1 条：4a68e02d3d695cee3c39d255ea5d20f7f2bf9b8c|claude-opus-5|pt-v1"
    },
    {
      "id": "MH5",
      "ok": true,
      "detail": "独立重算的 ds 读数必须与 noise-floor 产物逐 run 对上（§3.7 双向夹逼）—— 不一致 0 条"
    },
    {
      "id": "MH6",
      "ok": true,
      "detail": "4 条预注册预测的 sourceQuote 必须逐字命中对应 task JSON 的 expectedDiscrimination —— 未命中 0 条"
    },
    {
      "id": "MH7",
      "ok": true,
      "detail": "GT-001 曝光行必须复现 step11 §2.3.1 具名事实（popupbox-for-confirm=4 run，其余 4 条=0）—— 实测 alert-confirmText=0 handrolled-nav=0 hardcoded-color=0 popupbox-for-confirm=4 unconstrained-svg=0"
    },
    {
      "id": "MN1",
      "ok": true,
      "detail": "⛔ 二值口径不得出现任何 sd 派生字段（AGENTS §2.10，同 first-shot 的 MN3）—— 泄漏 0 处"
    },
    {
      "id": "MN2",
      "ok": true,
      "detail": "任一 task 的 N < 5 时全部判定字段必须为 null（fail closed，first-shot §2.5.1 那个坑）—— 全部达标（{\"GT-001\":5,\"GT-002\":5,\"GT-003\":5}）⇒ 允许出判定"
    },
    {
      "id": "MN3",
      "ok": true,
      "detail": "⛔ 组内噪声必须分 task 算，不得 pooled 成单一地板（尊重 noise-floor/first-shot/convergence 三个量具的断言）—— 实测每组 n = GT-001:5 GT-002:5 GT-003:5"
    },
    {
      "id": "MN4",
      "ok": true,
      "detail": "⛔ 本器不得读 context-manifest（输入面是产物不是上下文面，lab:E32）—— 实测读了 22 个文件，命中 0 个"
    },
    {
      "id": "MN5",
      "ok": true,
      "detail": "⛔ 预注册预测的 hit/miss 不得进控制集（miss 是读数不是失败，边界 ⑥）—— 实测控制编号 [MH1,MH2,MH3,MH4,MH5,MH6,MH7,MN1,MN2,MN3,MN4]，无 P* 编号（本条自己是最后一条，故上列不含 MN5）"
    }
  ],
  "boundaries": [
    "⛔ 区分度不是质量：三条 task 可区分 ≠ 哪条更好",
    "⛔ 铺 task 买不到统计力：组内 N 仍是 5，任何单条 task 的噪声地板一格不动",
    "⚠️ escape-rate 那族引自 noise-floor 产物，本器未独立重算（ds 那族有）",
    "⚠️ MDE 是功效分析的量不是检验统计量 ⇒ discriminable 以主检验 p 为准"
  ]
}