| Benchmark | Model | Rank | Results |
|---|---|---|---|
| arithmetic-reasoning-on-gsm8k | Qwen3.5-9B (rho=5%) | #2 | Accuracy: 97.7 |
| mathematical-reasoning-on-math-500 | Qwen3.5-9B (rho=30%) | #4 | Accuracy: 98.6 |
| mathematical-reasoning-on-aime24 | Qwen3.5-9B (rho=30%) | #6 | Acc: 93.3 |
| mathematical-reasoning-on-aime-2025 | Qwen3.5-9B (rho=30%) | #12 | Accuracy: 93.3 |
| mathematical-reasoning-on-aime-2026 | Qwen3.5-9B (rho=30%) | #19 | Accuracy: 90.0 |