README
Browse files
README.md
ADDED
|
@@ -0,0 +1,15 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
language: [ko]
|
| 4 |
+
pretty_name: chatbot-ft evaluation records
|
| 5 |
+
---
|
| 6 |
+
# chatbot-ft 평가 기록
|
| 7 |
+
|
| 8 |
+
[chatbot-ft](https://github.com/bigglz/chatbot-ft) 의 벤치마크 결과와 LLM-as-a-judge 기록. `scripts/hf_store.py` 가 자동으로 동기화한다.
|
| 9 |
+
|
| 10 |
+
```
|
| 11 |
+
benchmarks/<모델>/ run_benchmark.py 결과 (summary.md/json, raw/, logs/)
|
| 12 |
+
judge/conversations/<모델>/<생성조건 해시>/<실행ID>.jsonl 모델별 대화 (generation_config.json 의 조건이 같으면 상대 모델과 무관하게 재사용)
|
| 13 |
+
judge/pairs/<A>_vs_<B>/judgments_<평가자>/<실행ID>.jsonl 판정 기록 (순서를 바꿔 2회)
|
| 14 |
+
judge/pairs/<A>_vs_<B>/summary_<평가자>.md/json, review_<평가자>/ 집계·정성 검토 문서
|
| 15 |
+
```
|