Mathematical Reasoning on IMOBench
25.9Pass@1Composition-RL
Evaluation results
| Method | Links | |
|---|---|---|
| Composition-RLBackbone=Qwen3-14B-Base2026.02 | 25.9 | |
| Composition-RLBackbone=Qwen3-4B-Base, Curriculum=Depth 1 + 2 + 32026.02 | 22.9 | |
| Composition-RLBackbone=Qwen3-30B-A3B-Base2026.02 | 22.8 | |
| Standard RLVRBackbone=Qwen3-14B-Base2026.02 | 21.3 | |
| Composition-RLBackbone=Qwen3-4B-Base, Curriculum=Depth 1 + 22026.02 | 20.1 | |
| Composition-RLBackbone=Qwen3-8B-Base2026.02 | 18.4 | |
| Standard RLVRBackbone=Qwen3-8B-Base2026.02 | 16.2 | |
| Standard RLVRBackbone=Qwen3-4B-Base2026.02 | 14.4 | |
| Composition-RLBackbone=Qwen3-4B-Base2026.02 | 14.3 | |
| Standard RLVRBackbone=Qwen3-30B-A3B-Base2026.02 | 13.2 |