Mathematical Reasoning on Overall Aggregated
52.3Pass@1GRPO-VPS
Evaluation results
| Method | Links | ||
|---|---|---|---|
| GRPO-VPSBackbone=Qwen2.5-Math-7B, Supervision Type=With Process Supervision2026.04 | 52.3 | 3,121 | |
| GRPO w/ Skywork-7BBackbone=Qwen2.5-Math-7B, Supervision Type=With Process Supervision2026.04 | 52 | 3,791 | |
| GRPOBackbone=Qwen2.5-Math-7B, Supervision Type=Outcome Supervision Only2026.04 | 51.2 | 3,620 | |
| DrGRPOBackbone=Qwen2.5-Math-7B, Supervision Type=Outcome Supervision Only2026.04 | 51 | 3,534 | |
| GSPOBackbone=Qwen2.5-Math-7B, Supervision Type=Outcome Supervision Only2026.04 | 50.9 | 3,515 | |
| Eurus-2-7B-PRIMEBackbone=Qwen2.5-Math-7B, Supervision Type=With Process Supervision2026.04 | 50.1 | 4,382 | |
| S-GRPOBackbone=Qwen2.5-Math-7B, Supervision Type=With Process Supervision2026.04 | 49.6 | 3,136 | |
| GRPO-VPSBackbone=Qwen2.5-Math-1.5B, Supervision Type=With Process Supervision2026.04 | 43.6 | 3,355 | |
| DrGRPOBackbone=Qwen2.5-Math-1.5B, Supervision Type=Outcome Supervision Only2026.04 | 41.4 | 3,756 | |
| GRPOBackbone=Qwen2.5-Math-1.5B, Supervision Type=Outcome Supervision Only2026.04 | 41 | 3,772 | |
| GRPO w/ Skywork-1.5BBackbone=Qwen2.5-Math-1.5B, Supervision Type=With Process Supervision2026.04 | 40.7 | 3,607 | |
| S-GRPOBackbone=Qwen2.5-Math-1.5B, Supervision Type=With Process Supervision2026.04 | 39.1 | 3,523 | |
| GSPOBackbone=Qwen2.5-Math-1.5B, Supervision Type=Outcome Supervision Only2026.04 | 38.8 | 3,924 | |
| BASEBackbone=Qwen2.5-Math-7B, Supervision Type=Outcome Supervision Only2026.04 | 21.3 | 4,736 | |
| BASEBackbone=Qwen2.5-Math-1.5B, Supervision Type=Outcome Supervision Only2026.04 | 15.9 | 5,237 | |
| GRPO-VPSBackbone=Gemma-2-2B-it2026.04 | 11.7 | 2,399 | |
| GRPOBackbone=Gemma-2-2B-it2026.04 | 11.1 | 3,115 | |
| BASEBackbone=Gemma-2-2B-it2026.04 | 6.9 | 391 |