Instruction Following on AlpacaEval against gpt4-1106-preview (test)
31.3Win Rate (Raw)DIR
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DIRBase Model=Meta Llama3.1-8B-Instruct2025.12 | 31.3 | 19.66 | |
| SkyworkBase Model=Meta Llama3.1-8B-Instruct2025.12 | 29.38 | 13.21 | |
| ALBMBase Model=Meta Llama3.1-8B-Instruct2025.12 | 26.83 | 10.61 | |
| PoEBase Model=Meta Llama3.1-8B-Instruct2025.12 | 26.58 | 11.41 | |
| InfoRMBase Model=Meta Llama3.1-8B-Instruct2025.12 | 25.22 | 11.02 | |
| SkyworkBase Model=OpenRLHF Llama-3-8B-SFT2025.12 | 10.19 | 3.93 | |
| DIRBase Model=OpenRLHF Llama-3-8B-SFT2025.12 | 9.5 | 5.46 | |
| ALBMBase Model=OpenRLHF Llama-3-8B-SFT2025.12 | 8.88 | 5.08 | |
| PoEBase Model=OpenRLHF Llama-3-8B-SFT2025.12 | 7.14 | 3.28 | |
| InfoRMBase Model=OpenRLHF Llama-3-8B-SFT2025.12 | 5.84 | 3.65 |