Mathematical Reasoning on PolyMath medium (EN) 1.0 (test)
32AccuracyMAPO
Evaluation results
| Method | Links | |
|---|---|---|
| MAPOBackbone=Qwen2.5-7B-Instruct2026.04 | 32 | |
| polyGRPOBackbone=Qwen2.5-7B-Instruct2026.04 | 29.6 | |
| xRFTBackbone=Qwen2.5-7B-Instruct2026.04 | 28.8 | |
| LIDRBackbone=Qwen2.5-7B-Instruct2026.04 | 28 | |
| BaseBackbone=Qwen2.5-7B-Instruct2026.04 | 26.4 | |
| GRPOBackbone=Qwen2.5-7B-Instruct2026.04 | 26.4 | |
| polyGRPOBackbone=Llama3-8B-Instruct2026.04 | 9.6 | |
| BaseBackbone=Llama3-8B-Instruct2026.04 | 8 | |
| GRPOBackbone=Llama3-8B-Instruct2026.04 | 8 | |
| LIDRBackbone=Llama3-8B-Instruct2026.04 | 7.2 | |
| xRFTBackbone=Llama3-8B-Instruct2026.04 | 6.4 | |
| MAPOBackbone=Llama3-8B-Instruct2026.04 | 6.4 |