Graduate-level Professional Science Reasoning on GPQA
32.8Genetics AccuracyHIPPO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| HIPPOTraining Paradigm=Ours2026.06 | 32.8 | 48.8 | |
| SP3FTraining Paradigm=Reward-Shaping2026.06 | 28 | 45.6 | |
| Qwen2.5_7BTraining Paradigm=Standard Training, Configuration=+ SFT2026.06 | 26.6 | 44.4 | |
| Qwen2.5_7BTraining Paradigm=Standard Training, Configuration=+ RL2026.06 | 25 | 44.8 | |
| PREF-GRPOTraining Paradigm=Reward-Shaping2026.06 | 25 | 41.1 | |
| Qwen2.5_7BTraining Paradigm=Standard Training2026.06 | 21.9 | 47.9 | |
| PRM RLTraining Paradigm=Reward-Shaping2026.06 | 21.8 | 48.7 |