{
  "schema": 1,
  "scope": [
    "以下流程与样例由本站设计；样例均为虚构教学输入，参考文件不是产品运行记录。",
    "Workflows and samples are designed for this site. Inputs are fictional teaching material; references are not product execution records."
  ],
  "fictional": true,
  "change": "rows.filter(passed !== false) over all rows -> only boolean-judged rows enter the numerator and the denominator",
  "wrong_outputs": [
    0.5,
    0.75,
    1
  ],
  "fixed_outputs": [
    0.5,
    0.5,
    null
  ],
  "denominator_rule": {
    "unjudged": "excluded from the denominator and reported as a separate count",
    "error_or_timeout": "excluded, with a reason code recorded per row",
    "empty_or_all_unjudged": "not computable; report no percentage"
  },
  "report_counts_with": "every percentage is paired with its graded and unjudged counts",
  "execution_status": "not_executed",
  "unknown": [
    "real dataset version",
    "model identifier and parameters",
    "prompt revision",
    "provider pricing and quota",
    "judge-model bias"
  ],
  "limit": "Case one does not distinguish the two functions. Regression thresholds, judge-model bias and dataset leakage checks are outside this exercise.",
  "rollback": "Restore score-bug.mjs or reverse the two-line scoring change.",
  "conclusion": "A metric number without a denominator rule is not an evaluation result."
}