Self-Preference Bias Analysis on AlpacaEval
74.1PIRLongCat-Flash-Chat
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| LongCat-Flash-Chat2026.04 | 74.1 | 43.4 | 0.307 | 1,311 | 890 | — | — | — | — | |
| Gemma-3-12B2026.04 | 59.8 | 41.8 | 0.181 | 1,305 | 864 | — | — | — | — | |
| Gemma-3-27B2026.04 | 56.8 | 41.6 | 0.152 | 1,357 | 902 | — | — | — | — | |
| DeepSeek-V3.22026.04 | 56.6 | 34.1 | 0.226 | 1,349 | 916 | — | — | — | — | |
| GLM-4.5-Air2026.04 | 51 | 41.5 | 0.095 | 1,336 | 827 | — | — | — | — | |
| Qwen3-235B-A22B-25072026.04 | 50.4 | 41.5 | 0.09 | 1,396 | 948 | — | — | — | — | |
| Grok-3-Mini2026.04 | 49.9 | 39.9 | 0.1 | 1,362 | 902 | — | — | — | — | |
| Qwen3-235B-A22B-Thinking-25072026.04 | 47.6 | 35.2 | 0.124 | 1,228 | 554 | — | — | — | — | |
| DeepSeek-V3-03242026.04 | 39.5 | 37.1 | 0.024 | 1,368 | 897 | — | — | — | — | |
| Grok-4-Fast2026.04 | 37.3 | 33.7 | 0.035 | 1,371 | 922 | — | — | — | — | |
| Kimi-Linear-48B-A3B-Instruct2026.04 | 31.4 | 35.7 | -0.043 | 1,283 | 846 | — | — | — | — | |
| Hunyuan-A13B-Instruct2026.04 | 23.6 | 38.8 | -0.152 | 1,098 | 714 | — | — | — | — | |
| Claude-Sonnet-4.52026.04 | 19.3 | 42.2 | -0.229 | 1,155 | 773 | — | — | — | — | |
| Kimi-Dev-72B2026.04 | 12.5 | 24.2 | -0.117 | 1,106 | 753 | — | — | — | — | |
| Llama-3.3-70B-Instruct2026.04 | 11.9 | 27.1 | -0.151 | 1,331 | 898 | — | — | — | — | |
| Mistral-Nemo2026.04 | 3 | 8.2 | -0.052 | 1,183 | 766 | — | — | — | — | |
| Llama-3.1-8B-Instruct2026.04 | 2 | 8 | -0.06 | 1,050 | 710 | — | — | — | — | |
| DeepSeek-R1-05282026.04 | 0.8 | 10.5 | -0.097 | 1,247 | 845 | — | — | — | — | |
| Llama-3.2-3B-Instruct2026.04 | 0.7 | 0.7 | -0.001 | 1,004 | 667 | — | — | — | — | |
| Kimi-K2-Thinking2026.04 | 0.7 | 10.8 | -0.102 | 1,296 | 848 | — | — | — | — |