Mathematical Reasoning on AMC 24 (Accuracy)
47.4Accuracy (AMC 24)IBTPO
Evaluation Results
| Method | Links | |
|---|---|---|
| IBTPOBackbone=Qwen3-14B-Base2026.05 | 47.4 | |
| IBTPOMaximum truncation length=4K2026.05 | 46.2 | |
| IBTPOMaximum truncation length=8K2026.05 | 46.1 | |
| GRPO w/ Entropy RegBackbone=Qwen3-14B-Base2026.05 | 45.4 | |
| IBROBackbone=Qwen3-14B-Base2026.05 | 45.4 | |
| TreeRLBackbone=Qwen3-14B-Base2026.05 | 43.6 | |
| TreeRLMaximum truncation length=8K2026.05 | 40.5 | |
| Vanilla GRPOBackbone=Qwen3-14B-Base2026.05 | 40.4 | |
| TreeRLMaximum truncation length=4K2026.05 | 40.3 | |
| Vanilla GRPOMaximum truncation length=8K2026.05 | 39.8 | |
| Vanilla GRPOMaximum truncation length=4K2026.05 | 39.6 | |
| Initial ModelBackbone=Qwen3-14B-Base2026.05 | 30.7 | |
| Initial ModelMaximum truncation length=8K2026.05 | 26.9 | |
| Initial ModelMaximum truncation length=4K2026.05 | 26.7 |