Mathematics on MATH 500 (Accuracy)
96.4AccuracyDeepSeek-R1
Evaluation Results
| Method | Links | |
|---|---|---|
| DeepSeek-R12025.08 | 96.4 | |
| DS-32B-MathDomain=Math2025.08 | 95.2 | |
| NTele-R1-32B2025.08 | 95.2 | |
| DS-32B-CodeDomain=Code2025.08 | 94.8 | |
| QwQ-32B2025.08 | 94.6 | |
| DS-32B2025.08 | 89.8 | |
| MARFTBackbone=LaMAS, Scale=Quad2025.04 | 49.84 | |
| IPPO (MAPoRL)Backbone=LaMAS, Scale=Quad2025.04 | 49.76 | |
| MARFTBackbone=LaMAS, Scale=Trio2025.04 | 49.52 | |
| IPPO (MAPoRL)Backbone=LaMAS, Scale=Duo2025.04 | 49.48 | |
| MARFTBackbone=LaMAS, Scale=Duo2025.04 | 49.48 | |
| IPPO (MAPoRL)Backbone=LaMAS, Scale=Trio2025.04 | 49.12 | |
| VanillaBackbone=LaMAS, Scale=Quad2025.04 | 46.64 | |
| VanillaBackbone=LaMAS, Scale=Trio2025.04 | 45.76 | |
| VanillaBackbone=LaMAS, Scale=Duo2025.04 | 43.2 |