Scientific Question Answering on GPQA (Score)
33.48ScoreLARFT
Evaluation Results
| Method | Links | |
|---|---|---|
| LARFTBase Model=Qwen2.5-7B-Instruct2026.02 | 33.48 | |
| PositionIDBase Model=Qwen2.5-7B-Instruct2026.02 | 33.26 | |
| OriginBase Model=Qwen2.5-7B-Instruct2026.02 | 33.04 | |
| SFTBase Model=Llama-3.1-8B-Instruct2026.02 | 32.81 | |
| RLBase Model=Qwen2.5-7B-Instruct2026.02 | 32.59 | |
| OriginBase Model=Llama-3.2-3B-Instruct2026.02 | 32.37 | |
| SFTBase Model=Qwen2.5-7B-Instruct2026.02 | 32.14 | |
| SFT+RLBase Model=Qwen2.5-7B-Instruct2026.02 | 32.14 | |
| RLBase Model=Llama-3.1-8B-Instruct2026.02 | 32.14 | |
| RulerBase Model=Qwen2.5-7B-Instruct2026.02 | 31.7 | |
| RLBase Model=Llama-3.2-3B-Instruct2026.02 | 31.47 | |
| SFTBase Model=Llama-3.2-3B-Instruct2026.02 | 31.25 | |
| OriginBase Model=Llama-3.1-8B-Instruct2026.02 | 31.25 | |
| LARFTBase Model=Llama-3.1-8B-Instruct2026.02 | 31.25 | |
| OriginBase Model=Qwen2.5-3B-Instruct2026.02 | 31.03 | |
| LARFTBase Model=Llama-3.2-3B-Instruct2026.02 | 30.58 | |
| RLBase Model=Qwen2.5-3B-Instruct2026.02 | 30.13 | |
| SFT+RLBase Model=Llama-3.2-3B-Instruct2026.02 | 29.91 | |
| SFT+RLBase Model=Qwen2.5-3B-Instruct2026.02 | 29.46 | |
| SFTBase Model=Qwen2.5-3B-Instruct2026.02 | 29.24 | |
| LARFTBase Model=Qwen2.5-3B-Instruct2026.02 | 29.01 | |
| PositionIDBase Model=Qwen2.5-3B-Instruct2026.02 | 28.35 | |
| SFT+RLBase Model=Llama-3.1-8B-Instruct2026.02 | 28.35 | |
| RulerBase Model=Qwen2.5-3B-Instruct2026.02 | 27.9 | |
| PositionIDBase Model=Llama-3.1-8B-Instruct2026.02 | 25.45 | |
| PositionIDBase Model=Llama-3.2-3B-Instruct2026.02 | 24.33 | |
| RulerBase Model=Llama-3.2-3B-Instruct2026.02 | 24.33 | |
| RulerBase Model=Llama-3.1-8B-Instruct2026.02 | 24.33 |