| Benchmark | Model | Rank | Results |
|---|---|---|---|
| arithmetic-reasoning-on-gsm8k | Qwen3-8B with MixSD (lambda=0.5) | – | Accuracy: 93.3 |
| code-generation-on-humaneval | Qwen3-8B with MixSD (lambda=0.3) | – | Pass@1: 86 |
| mathematical-reasoning-on-aime24 | Qwen3-8B with MixSD (lambda=0.7) | – | Acc: 32.1 |
| mathematical-reasoning-on-math-500 | Qwen3-8B with MixSD (lambda=0.7) | – | Accuracy: 85.4 |