| Benchmark | Model | Rank | Results |
|---|---|---|---|
| mathematical-reasoning-on-aime-2025 | DAPO+HA-DW (Qwen-3-8B-Base) | – | Accuracy: 23.3 |
| mathematical-reasoning-on-math-500 | DAPO+HA-DW (Qwen-3-8B-Base) | – | Accuracy: 82.8 |
| mathematical-reasoning-on-minerva-math | GSPO+HA-DW (Qwen-3-8B-Base) | – | Accuracy: 41.9 |
| mathematical-reasoning-on-olympiadbench | DAPO+HA-DW (Qwen-3-8B-Base) | – | Accuracy: 50.0 |