Scientific Reasoning on ChemBench (test)
66Initial ScoreQWEN3-MAX
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| QWEN3-MAXModel Scale=Max2026.05 | 66 | 1 | 1.5 | 0.5 | 67 | |
| GRPOModel Scale=9B2026.05 | 63 | -2.5 | 0 | 2.5 | 60.5 | |
| DPOModel Scale=9B2026.05 | 61.5 | 0 | 3.5 | 3.5 | 61.5 | |
| ReCritModel Scale=9B2026.05 | 61.5 | 8 | 12.5 | 4.5 | 69.5 | |
| CRITIQUEGRPOModel Scale=9B2026.05 | 60.5 | 2 | 4 | 2 | 62.5 | |
| BASE MODELModel Scale=9B2026.05 | 59.5 | 0.5 | 3 | 2.5 | 60 | |
| SFTModel Scale=9B2026.05 | 59.5 | 1 | 4 | 3 | 60.5 | |
| GRPOModel Scale=4B2026.05 | 56 | -1.5 | 1 | 2.5 | 54.5 | |
| SFTModel Scale=4B2026.05 | 55.5 | -0.5 | 1.5 | 2 | 55 | |
| DPOModel Scale=4B2026.05 | 55 | 0.5 | 2.5 | 2 | 55.5 | |
| BASE MODELModel Scale=4B2026.05 | 53.5 | -0.5 | 1 | 1.5 | 53 | |
| CRITIQUEGRPOModel Scale=4B2026.05 | 52.5 | 1.5 | 4 | 2.5 | 54 | |
| ReCritModel Scale=4B2026.05 | 50.5 | 10.5 | 16.5 | 6 | 61 |