| Benchmark | Model | Rank | Results |
|---|---|---|---|
| arithmetic-reasoning-on-gsm8k | GRPO (with CAPO) (Qwen2.5-7B-Math) | – | Accuracy: 88.9 |
| mathematical-reasoning-on-aime24 | RLOO (with CAPO) (Qwen2.5-7B-Math) | – | Acc: 33.3 |
| mathematical-reasoning-on-math-500 | GRPO (with CAPO) (Qwen2.5-7B-Math) | – | Accuracy: 76.8 |
| mathematical-reasoning-on-minerva-math | REINFORCE++ (with CAPO) (Qwen2.5-7B-Math) | – | Accuracy: 40.1 |
| mathematical-reasoning-on-olympiadbench | GRPO (with CAPO) (Qwen2.5-7B-Math) | – | Accuracy: 39.7 |