Instruction Following on AlpacaEval Chinese
70.4Win RateCLO
Evaluation results
| Method | Links | ||
|---|---|---|---|
| CLOModel=Llama-3-8B, Eval Lang=Target2025.05 | 70.4 | 11.3 | |
| CLOModel=Llama-2-13B, Eval Lang=Target2025.05 | 65.2 | 5.9 | |
| CLOModel=Llama-3-8B, Eval Lang=English2025.05 | 64 | 12 | |
| CLOModel=Llama-2-7B, Eval Lang=Target2025.05 | 61.1 | 2 | |
| CLOModel=Qwen-2.5-3B, Eval Lang=Target2025.05 | 59.9 | 5.7 | |
| SFT+DPOModel=Llama-2-13B, Eval Lang=Target2025.05 | 59.3 | — | |
| SFT+DPOModel=Llama-3-8B, Eval Lang=Target2025.05 | 59.1 | — | |
| SFT+DPOModel=Llama-2-7B, Eval Lang=Target2025.05 | 59.1 | — | |
| CLOModel=Mistral-7B-v0.1, Eval Lang=Target2025.05 | 57.4 | 0.6 | |
| CLOModel=Mistral-7B-v0.1, Eval Lang=English2025.05 | 57.1 | 5.7 | |
| SFT+DPOModel=Mistral-7B-v0.1, Eval Lang=Target2025.05 | 56.8 | — | |
| CLOModel=Qwen-2.5-3B, Eval Lang=English2025.05 | 56.7 | 6 | |
| CLOModel=Llama-2-7B, Eval Lang=English2025.05 | 55.7 | 4.6 | |
| SFT+DPOModel=Qwen-2.5-3B, Eval Lang=Target2025.05 | 54.2 | — | |
| SFT+DPOModel=Llama-3-8B, Eval Lang=English2025.05 | 52 | — | |
| SFT+DPOModel=Mistral-7B-v0.1, Eval Lang=English2025.05 | 51.4 | — | |
| SFT+DPOModel=Llama-2-7B, Eval Lang=English2025.05 | 51.1 | — | |
| SFT+DPOModel=Qwen-2.5-3B, Eval Lang=English2025.05 | 50.7 | — | |
| SFT+DPOModel=Llama-2-13B, Eval Lang=English2025.05 | 50.6 | — | |
| CLOModel=Llama-2-13B, Eval Lang=English2025.05 | 50.4 | -0.2 |