Scientific Reasoning on Science Q&A (test)
76.02Chemistry AccuracyPGPO (best alpha)
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| PGPO (best alpha)Alpha sensitivity parameter (α)=best, evaluation protocol=avg@16, initial checkpoint=Qwen3-8B2026.06 | 76.02 | 77.56 | 61.25 | 78.65 | |
| SDPOevaluation protocol=avg@16, initial checkpoint=Qwen3-8B2026.06 | 72.57 | 77.05 | 61.62 | 74.14 |