Mathematical Reasoning on PolyMath medium ALL 1.0 (test)
25.24AccuracypolyGRPO
Evaluation results
| Method | Links | |
|---|---|---|
| polyGRPOBackbone=Qwen2.5-7B-Instruct2026.04 | 25.24 | |
| LIDRBackbone=Qwen2.5-7B-Instruct2026.04 | 25.07 | |
| BaseBackbone=Qwen2.5-7B-Instruct2026.04 | 23.69 | |
| GRPOBackbone=Qwen2.5-7B-Instruct2026.04 | 23.42 | |
| MAPOBackbone=Qwen2.5-7B-Instruct2026.04 | 23.24 | |
| xRFTBackbone=Qwen2.5-7B-Instruct2026.04 | 18.4 | |
| polyGRPOBackbone=Llama3-8B-Instruct2026.04 | 5.51 | |
| LIDRBackbone=Llama3-8B-Instruct2026.04 | 5.02 | |
| xRFTBackbone=Llama3-8B-Instruct2026.04 | 4.93 | |
| BaseBackbone=Llama3-8B-Instruct2026.04 | 4.84 | |
| GRPOBackbone=Llama3-8B-Instruct2026.04 | 4.76 | |
| MAPOBackbone=Llama3-8B-Instruct2026.04 | 4.31 |