| Benchmark | Model | Rank | Results |
|---|---|---|---|
| mathematical-reasoning-on-minerva-math | Qwen2.5-Math-7B with TraPO (4K labels, 12K unlabeled) | #24 | Accuracy: 39.3 |
| mathematical-reasoning-on-math-500 | Qwen2.5-Math-7B with TraPO (4K labels, 12K unlabeled) | #74 | Accuracy: 84.6 |
| multi-task-language-understanding-on-mmlu-pro | Qwen2.5-Math-7B with TraPO (4K labels, 12K unlabeled) | #114 | Accuracy: 50.7 |