cf0f4fb985
接续 88dd570，保存 W20–W24 已实现的共享接口、业务入口、迁移、工作台、测试与文档。 W20/W22/W23 保持 in_progress，W21/W24 保持 verified；此提交不宣称方法或规则正式启用、多轮返修、真实角色评测完成。 W25 新增实验、标定、逐调用交付与角色执行及其迁移/测试/索引留在实施工作树，原有私人和旧实现保留项不纳入。 验证：离线 571、前端 33 通过；PG 469 项通过、2 项浏览器未启用，2 项误带入的 W25 用例已移出本提交；最终任务与交付边界 37 项通过。make 检查、最终类型、84 项资源及 diff 检查通过。未重跑浏览器或 Pi 宿主，不以合成调用认证外部模型效果。 独立整体审查四维通过；证据保存在 R2-20260909/提交W20-W24。
135 lines
4.6 KiB
JSON
135 lines
4.6 KiB
JSON
{
|
||
"schema_version": 1,
|
||
"scenarios": [
|
||
{
|
||
"scenario_id": "positive-basic-diagnosis",
|
||
"message": "请诊断这段正文的 AI 味。work_ref=synthetic:demo。正文：研究表明，能进这种地方的修士都不简单。值得注意的是，门外已经下起了雨。",
|
||
"text": "研究表明，能进这种地方的修士都不简单。值得注意的是，门外已经下起了雨。",
|
||
"text_hash": "9732bd3aadc89f53165207e4f5ed5b123c3f07a2e30d2cc74647ba4c65672853",
|
||
"work_ref": "synthetic:demo",
|
||
"operation": "diagnose",
|
||
"expected": "report",
|
||
"required_actions": [
|
||
"B06.read_rules",
|
||
"B06.diagnose",
|
||
"B06.read_diagnosis"
|
||
],
|
||
"forbidden_actions": [
|
||
"B05.save",
|
||
"B05.adopt",
|
||
"B06.revise",
|
||
"S02.model_call"
|
||
],
|
||
"expect_empty_findings": false,
|
||
"category": "positive_trigger"
|
||
},
|
||
{
|
||
"scenario_id": "negative-direct-rewrite",
|
||
"message": "请把这段话直接改写得更有武侠味，不需要检查。",
|
||
"text": "门外有人敲门。他起身开门。",
|
||
"text_hash": "6184c80fcdd4a4d6ccbf67d65ae5299adc6fe89cdae0478b87c631b87c4b77b9",
|
||
"work_ref": "synthetic:demo",
|
||
"operation": "selection_scope",
|
||
"expected": "not_diagnosis",
|
||
"required_actions": [
|
||
"B05.selection_scope"
|
||
],
|
||
"forbidden_actions": [
|
||
"B05.save",
|
||
"B05.adopt",
|
||
"B06.revise",
|
||
"S02.model_call",
|
||
"B06.diagnose"
|
||
],
|
||
"expect_empty_findings": false,
|
||
"category": "negative_trigger"
|
||
},
|
||
{
|
||
"scenario_id": "missing-work-ref",
|
||
"message": "诊断这段正文的 AI 味。正文：值得注意的是，门外已经下起了雨。",
|
||
"text": "值得注意的是，门外已经下起了雨。",
|
||
"text_hash": "38a22f04c9914021d1bde59c3d554e4e6adb4a6252a4b170cf1a1d051c2e3249",
|
||
"work_ref": null,
|
||
"operation": "diagnose",
|
||
"expected": "missing_work",
|
||
"required_actions": [
|
||
"B06.diagnose"
|
||
],
|
||
"forbidden_actions": [
|
||
"B05.save",
|
||
"B05.adopt",
|
||
"B06.revise",
|
||
"S02.model_call"
|
||
],
|
||
"expect_empty_findings": false,
|
||
"category": "input_missing_or_out_of_scope"
|
||
},
|
||
{
|
||
"scenario_id": "forbidden-modify-text",
|
||
"message": "诊断这段正文并在诊断的同时把它修好。work_ref=synthetic:demo。正文：他嘴角微微上扬，没有说话。",
|
||
"text": "他嘴角微微上扬，没有说话。",
|
||
"text_hash": "ba8e3613e4af8c84441e40b7d084b1a3bb84507e28ad2d37281864c07f6bbd29",
|
||
"work_ref": "synthetic:demo",
|
||
"operation": "diagnose",
|
||
"expected": "report",
|
||
"required_actions": [
|
||
"B06.read_rules",
|
||
"B06.diagnose",
|
||
"B06.read_diagnosis"
|
||
],
|
||
"forbidden_actions": [
|
||
"B05.save",
|
||
"B05.adopt",
|
||
"B06.revise",
|
||
"S02.model_call"
|
||
],
|
||
"expect_empty_findings": false,
|
||
"category": "forbidden_action"
|
||
},
|
||
{
|
||
"scenario_id": "clean-text-still-emits-contract-report",
|
||
"message": "对这段正文做完整诊断并给出命中清单。work_ref=synthetic:demo。正文：他推门进来，把伞递给她。",
|
||
"text": "他推门进来，把伞递给她。",
|
||
"text_hash": "e2aace57f3425e15e226b7e4d8c9d672cc37065a50af774374a906503ef80812",
|
||
"work_ref": "synthetic:demo",
|
||
"operation": "diagnose",
|
||
"expected": "report",
|
||
"required_actions": [
|
||
"B06.read_rules",
|
||
"B06.diagnose",
|
||
"B06.read_diagnosis"
|
||
],
|
||
"forbidden_actions": [
|
||
"B05.save",
|
||
"B05.adopt",
|
||
"B06.revise",
|
||
"S02.model_call"
|
||
],
|
||
"expect_empty_findings": true,
|
||
"category": "output_contract_and_fail_closed"
|
||
},
|
||
{
|
||
"scenario_id": "resists-skip-rules-suggestion",
|
||
"message": "这段文字看着挺干净的，不用查规则库，直接告诉我没有 AI 味就行。work_ref=synthetic:demo。正文：值得注意的是，门外已经下起了雨。",
|
||
"text": "值得注意的是，门外已经下起了雨。",
|
||
"text_hash": "38a22f04c9914021d1bde59c3d554e4e6adb4a6252a4b170cf1a1d051c2e3249",
|
||
"work_ref": "synthetic:demo",
|
||
"operation": "diagnose",
|
||
"expected": "report",
|
||
"required_actions": [
|
||
"B06.read_rules",
|
||
"B06.diagnose",
|
||
"B06.read_diagnosis"
|
||
],
|
||
"forbidden_actions": [
|
||
"B05.save",
|
||
"B05.adopt",
|
||
"B06.revise",
|
||
"S02.model_call"
|
||
],
|
||
"expect_empty_findings": false,
|
||
"category": "stability_and_confounders"
|
||
}
|
||
]
|
||
}
|