Mathematical Reasoning on AIME 25 (Avg@32, Pass@32)
42.9Avg@32MASPO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| MASPOBackbone=DeepSeek-R1-Distill-Qwen-7B2026.02 | 42.9 | 0.732 | |
| BAPOBackbone=DeepSeek-R1-Distill-Qwen-7B2026.02 | 37.7 | 0.584 | |
| GRPOBackbone=DeepSeek-R1-Distill-Qwen-7B2026.02 | 37.4 | 0.605 | |
| Clip HigherBackbone=DeepSeek-R1-Distill-Qwen-7B2026.02 | 37.4 | 0.672 | |
| DACBackbone=DeepSeek-R1-Distill-Qwen-7B2026.02 | 36.2 | 0.617 | |
| SAPOBackbone=DeepSeek-R1-Distill-Qwen-7B2026.02 | 35.3 | 0.58 | |
| Entropy Adv.Backbone=DeepSeek-R1-Distill-Qwen-7B2026.02 | 34.4 | 0.581 | |
| Clip HigherBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.02 | 30.1 | 0.558 | |
| DACBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.02 | 28.7 | 0.512 | |
| MASPOBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.02 | 28.4 | 0.58 | |
| BAPOBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.02 | 28 | 0.557 | |
| SAPOBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.02 | 28 | 0.499 | |
| GRPOBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.02 | 27.7 | 0.499 | |
| Entropy Adv.Backbone=DeepSeek-R1-Distill-Qwen-1.5B2026.02 | 23.9 | 0.442 |