Comprehensive Evaluation on Overall Across Benchmarks
44.3Avg@32 AccuracyIBTPO
Evaluation Results
| Method | Links | |
|---|---|---|
| IBTPOBase Model=Qwen3-8B-Base2026.05 | 44.3 | |
| TreeRLBase Model=Qwen3-8B-Base2026.05 | 42 | |
| IBROBase Model=Qwen3-8B-Base2026.05 | 41.6 | |
| TreePOBase Model=Qwen3-8B-Base2026.05 | 41.6 | |
| GRPO w/ Entropy Reg†Base Model=Qwen3-8B-Base2026.05 | 41.4 | |
| GRPO w/ Clip-higherBase Model=Qwen3-8B-Base2026.05 | 41 | |
| Vanilla GRPOBase Model=Qwen3-8B-Base2026.05 | 40.7 | |
| Initial ModelBase Model=Qwen3-8B-Base2026.05 | 32 | |
| IBTPOBase Model=Qwen3-1.7B-Base2026.05 | 29.2 | |
| TreeRLBase Model=Qwen3-1.7B-Base2026.05 | 26.8 | |
| Vanilla GRPOBase Model=Qwen3-1.7B-Base2026.05 | 26.3 | |
| TreePOBase Model=Qwen3-1.7B-Base2026.05 | 26.2 | |
| GRPO w/ Clip-higherBase Model=Qwen3-1.7B-Base2026.05 | 26 | |
| IBROBase Model=Qwen3-1.7B-Base2026.05 | 25.9 | |
| GRPO w/ Entropy Reg†Base Model=Qwen3-1.7B-Base2026.05 | 24.9 | |
| Initial ModelBase Model=Qwen3-1.7B-Base2026.05 | 19.9 |