Mathematics on AMC 24 (avg@32 accuracy)
46Accuracy (avg@32)IBTPO
Evaluation Results
| Method | Links | |
|---|---|---|
| IBTPOBase Model=Qwen3-8B-Base2026.05 | 46 | |
| TreeRLBase Model=Qwen3-8B-Base2026.05 | 40.5 | |
| TreePOBase Model=Qwen3-8B-Base2026.05 | 40.1 | |
| IBROBase Model=Qwen3-8B-Base2026.05 | 39.5 | |
| Vanilla GRPOBase Model=Qwen3-8B-Base2026.05 | 39.4 | |
| GRPO w/ Entropy Reg†Base Model=Qwen3-8B-Base2026.05 | 39.4 | |
| GRPO w/ Clip-higherBase Model=Qwen3-8B-Base2026.05 | 38.9 | |
| Initial ModelBase Model=Qwen3-8B-Base2026.05 | 26.8 | |
| IBTPOBase Model=Qwen3-1.7B-Base2026.05 | 23.4 | |
| IBROBase Model=Qwen3-1.7B-Base2026.05 | 20.8 | |
| TreeRLBase Model=Qwen3-1.7B-Base2026.05 | 20.6 | |
| GRPO w/ Clip-higherBase Model=Qwen3-1.7B-Base2026.05 | 20.5 | |
| Vanilla GRPOBase Model=Qwen3-1.7B-Base2026.05 | 19.7 | |
| TreePOBase Model=Qwen3-1.7B-Base2026.05 | 19.5 | |
| GRPO w/ Entropy Reg†Base Model=Qwen3-1.7B-Base2026.05 | 18.8 | |
| Initial ModelBase Model=Qwen3-1.7B-Base2026.05 | 15.1 |