Scientific Reasoning on Scientific Methods Reasoning Internal (test)
70.851-σ ScoreDirect-SFT
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Direct-SFTModel=Qwen32026.04 | 70.85 | 85.56 | 84.62 | 77.1 | |
| Reason-DAPOModel=Qwen32026.04 | 69.43 | 75 | 83.52 | 72.73 | |
| Reason-SFT-DrGRPOModel=Qwen32026.04 | 66.19 | 76.41 | 83.52 | 71.35 | |
| Synthetic-Reason-SFT-DAPOModel=Qwen32026.04 | 65.79 | 72.53 | 74.72 | 68.93 | |
| Reason-SFT-DAPOModel=Qwen32026.04 | 64.57 | 79.23 | 83.52 | 71.35 | |
| Reasoning (high)Model=GPT-52026.04 | 61.94 | 61.27 | 56.04 | 61.1 | |
| Reasoning (med)Model=GPT-52026.04 | 59.1 | 61.62 | 56.04 | 59.61 | |
| Reasoning (low)Model=GPT-52026.04 | 58.7 | 58.45 | 49.45 | 57.65 | |
| Direct-SFTModel=Llama3.12026.04 | 53.64 | 58.1 | 67.03 | 56.5 | |
| Base (Reasoning)Model=Gemini 2.5 Flash2026.04 | 41.9 | 40.14 | 36.26 | 40.73 | |
| BaseModel=Llama3.12026.04 | 37.36 | 31.33 | 27.93 | 30.03 | |
| Reason-SFTModel=Qwen32026.04 | 35.32 | 38.9 | 45.05 | 37.51 | |
| Base (Reasoning)Model=Llama3.12026.04 | 26.52 | 30.63 | 21.98 | 27.39 | |
| BaseModel=Qwen32026.04 | 18.42 | 26.05 | 10.99 | 20.14 | |
| Base (Reasoning)Model=Qwen32026.04 | 15.38 | 27.11 | 26.11 | 25.31 |