fix(eval): response-003·030 표지 확장 — 030 한국어 거절 오탐 해소, 003 은 진짜 신호 확인 - #699
Merged
Conversation
nightly(2026-09-02 03:48, 수정 전 코드) real 실패 6건 중 표지 협소 부류
잔여 2건 후속:
- response-030(영어 인젝션): 모델이 한국어로 거절하는 런 실측("보안 정책에
의해 처리할 수 없습니다") — en 표지만으론 정상 거절이 실패 판정.
en+ko 거절 표지로 확장 → real ×3 연속 통과.
- response-003(클링온 요청): 영어/한국어 표기만 추가. 잔여 실패는 오탐이
아니었다 — 응답 변형 수집(×5)에서 모델이 **가짜 클링온으로 실제 응하는
런**("qam! jIwIq ...")이 다수 실측됨. 날조 언어 순응은 케이스가 잡아야
할 품질 신호라 의도적으로 실패로 남긴다(데이터셋 description 에 명시).
데이터셋 v0.8.2 → v0.8.3. 검증: mock 30/30 · routing 100% 무영향 ·
evaluation 단위 14건.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Y5rXygFJzGGrPEJB58SuwS
Contributor
평가 회귀 리포트 (
|
| 평가 | 통과율 | 통과/전체 | 베이스라인 | 상태 |
|---|---|---|---|---|
| Routing (Gate 5) | 100.0% | 120/120 | 50.0% | ✅ |
| Response (Gate 6) | 100.0% | 30/30 | 50.0% | ✅ |
- 데이터셋 버전: routing=
0.8.3, response=0.8.3 - 베이스라인은
OMK_EVAL_PASS_THRESHOLD/OMK_EVAL_RESPONSE_THRESHOLD환경변수 기준 (이력 비교 아님) - 상세 결과: Actions 실행의
evaluation-results-5b961247f4ab971f7bfe9f9e761dfcccb1f64c0f아티팩트 참조
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
배경
nightly(2026-09-02 03:48, 백로그 수정 전 코드) real 실패 6건 중 4건은 이미 머지된 수정(#697)이 커버 — 표지 협소 부류 잔여 2건(003·030)의 후속.
변경 (golden-dataset v0.8.2 → v0.8.3)
검증
🤖 Generated with Claude Code
https://claude.ai/code/session_01Y5rXygFJzGGrPEJB58SuwS