Pointwise Grading on AlignBench
0.997Pearson (r)GPT-4
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| GPT-4Level=System-Level, Setting=Reference-Free2023.11 | 0.997 | 0.976 | 0.929 | |
| GPT-4Level=System-Level, Setting=Referenced2023.11 | 0.995 | 1 | 1 | |
| CRITIQUELLMLevel=System-Level, Setting=Referenced2023.11 | 0.995 | 1 | 1 | |
| JudgeLM-13BLevel=System-Level, Setting=Referenced2023.11 | 0.984 | 0.976 | 0.929 | |
| Mixtral-8x7BLevel=System-Level, Setting=Referenced2023.11 | 0.972 | 0.976 | 0.929 | |
| ChatGPTLevel=System-Level, Setting=Referenced2023.11 | 0.955 | 0.976 | 0.929 | |
| CRITIQUELLMLevel=System-Level, Setting=Reference-Free2023.11 | 0.954 | 0.976 | 0.929 | |
| Qwen-14B-ChatLevel=System-Level, Setting=Referenced2023.11 | 0.901 | 0.929 | 0.786 | |
| Mixtral-8x7BLevel=System-Level, Setting=Reference-Free2023.11 | 0.863 | 0.929 | 0.786 | |
| Baichuan2-13B-ChatLevel=System-Level, Setting=Referenced2023.11 | 0.854 | 0.929 | 0.786 | |
| ChatGLM3-6BLevel=System-Level, Setting=Referenced2023.11 | 0.79 | 0.833 | 0.643 | |
| ChatGPTLevel=System-Level, Setting=Reference-Free2023.11 | 0.778 | 0.833 | 0.643 | |
| Qwen-14B-ChatLevel=System-Level, Setting=Reference-Free2023.11 | 0.772 | 0.833 | 0.643 | |
| JudgeLM-13BLevel=System-Level, Setting=Reference-Free2023.11 | 0.717 | 0.905 | 0.786 | |
| Baichuan2-13B-ChatLevel=System-Level, Setting=Reference-Free2023.11 | 0.663 | 0.527 | 0.4 | |
| Llama-2-70B-ChatLevel=System-Level, Setting=Referenced2023.11 | 0.663 | 0.667 | 0.5 | |
| GPT-4Level=Text-Level, Setting=Referenced2023.11 | 0.629 | 0.583 | 0.532 | |
| AUTO-J-Bilingual-6BLevel=System-Level, Setting=Reference-Free2023.11 | 0.558 | 0.571 | 0.5 | |
| CRITIQUELLMLevel=Text-Level, Setting=Referenced2023.11 | 0.555 | 0.523 | 0.477 | |
| Llama-2-70B-ChatLevel=System-Level, Setting=Reference-Free2023.11 | 0.547 | 0.429 | 0.286 | |
| ChatGLM3-6BLevel=System-Level, Setting=Reference-Free2023.11 | 0.544 | 0.548 | 0.429 | |
| GPT-4Level=Text-Level, Setting=Reference-Free2023.11 | 0.523 | 0.494 | 0.447 | |
| Mixtral-8x7BLevel=Text-Level, Setting=Referenced2023.11 | 0.474 | 0.471 | 0.426 | |
| JudgeLM-13BLevel=Text-Level, Setting=Referenced2023.11 | 0.45 | 0.43 | 0.391 | |
| ChatGPTLevel=Text-Level, Setting=Referenced2023.11 | 0.443 | 0.421 | 0.379 | |
| Qwen-14B-ChatLevel=Text-Level, Setting=Referenced2023.11 | 0.373 | 0.379 | 0.358 | |
| CRITIQUELLMLevel=Text-Level, Setting=Reference-Free2023.11 | 0.366 | 0.352 | 0.319 | |
| Mixtral-8x7BLevel=Text-Level, Setting=Reference-Free2023.11 | 0.302 | 0.306 | 0.282 | |
| ChatGPTLevel=Text-Level, Setting=Reference-Free2023.11 | 0.292 | 0.287 | 0.266 | |
| Qwen-14B-ChatLevel=Text-Level, Setting=Reference-Free2023.11 | 0.255 | 0.254 | 0.239 | |
| ChatGLM3-6BLevel=Text-Level, Setting=Referenced2023.11 | 0.223 | 0.222 | 0.207 | |
| Baichuan2-13B-ChatLevel=Text-Level, Setting=Referenced2023.11 | 0.199 | 0.2 | 0.187 | |
| JudgeLM-13BLevel=Text-Level, Setting=Reference-Free2023.11 | 0.17 | 0.162 | 0.155 | |
| ChatGLM3-6BLevel=Text-Level, Setting=Reference-Free2023.11 | 0.159 | 0.15 | 0.14 | |
| Llama-2-70B-ChatLevel=Text-Level, Setting=Referenced2023.11 | 0.152 | 0.162 | 0.109 | |
| Baichuan2-13B-ChatLevel=Text-Level, Setting=Reference-Free2023.11 | 0.125 | 0.117 | 0.11 | |
| Llama-2-70B-ChatLevel=Text-Level, Setting=Reference-Free2023.11 | 0.123 | 0.122 | 0.113 | |
| AUTO-J-Bilingual-6BLevel=Text-Level, Setting=Reference-Free2023.11 | 0.044 | 0.045 | 0.041 |