Reasoning on Reasoning Benchmarks Biology, Chemistry, Material, Physics, Tool Use
65.8Biology AccuracySRPO*
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| SRPO*Base Model=Qwen3-4B2026.06 | 65.8 | 82.7 | 81.3 | 74 | 67 | 74.2 | |
| DRIFTBase Model=Qwen3-4B2026.06 | 64.3 | 79.6 | 82.8 | 72.2 | 74 | 74.6 | |
| DRIFTBase Model=Olmo3-7B2026.06 | 61.3 | 82 | 79.2 | 71.3 | 70.2 | 72.8 | |
| SC-SDPO*Base Model=Olmo3-7B2026.06 | 55.9 | 81.5 | 79.1 | 71.1 | 62.9 | 70.1 | |
| GRPO*Base Model=Qwen3-4B2026.06 | 55.5 | 78.3 | 80.1 | 71.9 | 62.9 | 69.7 | |
| DistIL*Base Model=Olmo3-7B2026.06 | 55.3 | 81 | 76.9 | 74.5 | — | — | |
| SDPO*Base Model=Qwen3-4B2026.06 | 54 | 77.3 | 74.3 | 66.7 | 61.1 | 66.7 | |
| SDPO*Base Model=Olmo3-7B2026.06 | 52.8 | 80 | 79.1 | 66.1 | 62.1 | 68 | |
| GRPO*Base Model=Olmo3-7B2026.06 | 49.8 | 57.5 | 73.5 | 62.7 | 60.6 | 60.8 | |
| Qwen3-4BBase Model=Qwen3-4B, Configuration=Base model2026.06 | 30.8 | 43.6 | 61.2 | 59.8 | 57.9 | 50.7 | |
| Olmo3-7BBase Model=Olmo3-7B, Configuration=Base model2026.06 | 16.2 | 22.8 | 36.7 | 37.7 | 39.3 | 30.5 |