Quality-judgment accuracy on LLMBar LS
86.5Strict AccuracyDeepSeek
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| DeepSeekLanguage condition=LS, N=8382026.06 | 86.5 | 88.9 | 86.9 | 50.8 | 4.8 | |
| Gemini FlashLanguage condition=LS, N=8382026.06 | 82.3 | 82.9 | 80.4 | 59.2 | 1.2 | |
| Claude HaikuLanguage condition=LS, N=8382026.06 | 73.9 | 74.4 | 71.5 | 49.4 | 1.1 | |
| GPT-4.1 MiniLanguage condition=LS, N=8382026.06 | 72 | 73 | 70.1 | 50.8 | 2.1 |