| Benchmark | Model | Rank | Results |
|---|---|---|---|
| mathematical-reasoning-on-aime-2025 | KV4-RL (Qwen3-8B-Base) | – | Accuracy: 22.1 |
| mathematical-reasoning-on-aime24 | KV4-RL (Qwen3-8B-Base) | – | Acc: 29.7 |
| mathematical-reasoning-on-amc23 | KV4-RL (Qwen3-8B-Base) | – | Acc: 68.9 |
| mathematical-reasoning-on-math-500 | KV4-RL (Qwen3-8B-Base) | – | Accuracy: 87.7 |
| mathematical-reasoning-on-minerva-math | KV4-RL (Qwen3-8B-Base) | – | Accuracy: 51.8 |
| mathematical-reasoning-on-olympiadbench | KV4-RL (Qwen3-8B-Base) | – | Accuracy: 56.8 |