Mathematical Reasoning on Olympiad-Bench (Avg@32, Pass@32)
58Avg@32MASPO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| MASPOBackbone=DeepSeek-R1-Distill-Qwen-7B2026.02 | 58 | 74.9 | |
| DACBackbone=DeepSeek-R1-Distill-Qwen-7B2026.02 | 57.9 | 76 | |
| Clip HigherBackbone=DeepSeek-R1-Distill-Qwen-7B2026.02 | 57.5 | 75.5 | |
| GRPOBackbone=DeepSeek-R1-Distill-Qwen-7B2026.02 | 57.2 | 73.9 | |
| BAPOBackbone=DeepSeek-R1-Distill-Qwen-7B2026.02 | 57.1 | 74.5 | |
| SAPOBackbone=DeepSeek-R1-Distill-Qwen-7B2026.02 | 56.6 | 74.6 | |
| Entropy Adv.Backbone=DeepSeek-R1-Distill-Qwen-7B2026.02 | 56.4 | 73.7 | |
| SAPOBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.02 | 48.2 | 66.7 | |
| MASPOBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.02 | 47.8 | 65.7 | |
| Entropy Adv.Backbone=DeepSeek-R1-Distill-Qwen-1.5B2026.02 | 46.4 | 64.3 | |
| GRPOBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.02 | 46.3 | 64.7 | |
| DACBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.02 | 46.3 | 64.5 | |
| BAPOBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.02 | 44 | 61.3 | |
| Clip HigherBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.02 | 42 | 61.6 |