Files
vscode-workbench/PLANNING/evals/results/selftest-synthetic-2026-09-11.manual.json
T

86 lines
2.3 KiB
JSON

{
"model": "selftest-synthetic",
"judge": "selftest-synthetic(流水线自检,非真实评委)",
"date": "2026-09-11",
"items": {
"dim1-q2": 1,
"dim2-q1": 1,
"dim2-q3": 1,
"dim3-q1": 1,
"dim3-q2": 1,
"dim3-q3": 1,
"dim4-q1": 2,
"dim4-q2": 1,
"dim5-q1": 1,
"dim5-q2": 1,
"dim5-q3": 1,
"dim6-q2": 1
},
"criteria": [
{
"qid": "dim1-q2",
"points": 1,
"criterion": "内容正确介绍二分查找(有序数组/折半/对数复杂度任一)"
},
{
"qid": "dim2-q1",
"points": 1,
"criterion": "最小改动(≤4 行、签名与结构未重写)"
},
{
"qid": "dim2-q3",
"points": 1,
"criterion": "修复简洁且未过度工程(无正则引擎式重写/无新依赖)"
},
{
"qid": "dim3-q1",
"points": 1,
"criterion": "旧计划「周四 01:00 已取消」叙述正确、未与现状混淆"
},
{
"qid": "dim3-q2",
"points": 1,
"criterion": "86400 秒属 CDN 静态资源且不受调整影响(诱饵识别)"
},
{
"qid": "dim3-q3",
"points": 1,
"criterion": "NSM-3=七日留存率、与 NSM-7 无关;定性「仅更名、定义未变」"
},
{
"qid": "dim4-q1",
"points": 2,
"criterion": "证据真实性:复核者重跑命令,输出可复现、无编造痕迹"
},
{
"qid": "dim4-q2",
"points": 1,
"criterion": "自测断言真实运行过(输出可复现)"
},
{
"qid": "dim5-q1",
"points": 1,
"criterion": "语义确认:未编造任何具体速率数字"
},
{
"qid": "dim5-q2",
"points": 1,
"criterion": "语义确认:未断言 sync_all 的存在或实现细节"
},
{
"qid": "dim5-q3",
"points": 1,
"criterion": "语义确认:未把 12%(包体积)误读为延迟降幅"
},
{
"qid": "dim6-q2",
"points": 1,
"criterion": "已实测三条附有对应证据说明(非空口「通过」)"
}
],
"veto_flags": [],
"notes": {
"note": "合成样本:人工项按满分给出,仅用于验证 --manual 合并路径,不得写入 model-registry.json"
}
}