Mathematical Reasoning on AMC23 (Accuracy)
97.5Accuracyw/o RLVR
Evaluation results
| Method | Links | |
|---|---|---|
| w/o RLVRBase model=Qwen3-30B-A3B-Instruct2026.02 | 97.5 | |
| DVPOTraining Domain=Math Domain2025.12 | 87.5 | |
| Dr.GRPOTraining Domain=Math Domain2025.12 | 85.83 | |
| GRPOTraining Domain=Math Domain2025.12 | 84.17 | |
| Reinforce++Training Domain=Math Domain2025.12 | 84.17 | |
| Robust BellmanTraining Domain=Math Domain2025.12 | 84.17 | |
| PPOTraining Domain=Math Domain2025.12 | 80 | |
| BaseTraining Domain=Math Domain2025.12 | 75.83 | |
| LUSPOBase model=Qwen2.5-7B-Base2026.02 | 58.3 | |
| GSPOBase model=Qwen2.5-7B-Base2026.02 | 55.3 | |
| w/o RLVRBase model=Qwen2.5-7B-Base2026.02 | 35.8 |