Mathematics on Minerva Math
45.2Pass@1 AccuracyDPO
Evaluation Results
| Method | Links | |
|---|---|---|
| DPOBackbone=olmo-3.1-32b-it-sft2026.04 | 45.2 | |
| Pair+Backbone=olmo-3.1-32b-it-sft2026.04 | 44.5 | |
| Mag+Backbone=olmo-3.1-32b-it-sft2026.04 | 43.8 | |
| MPO+Backbone=olmo-3.1-32b-it-sft2026.04 | 43.8 | |
| Soft+Backbone=olmo-3.1-32b-it-sft2026.04 | 43.4 | |
| RFTBackbone=olmo-3.1-32b-it-sft2026.04 | 43 | |
| INSIGHTModel=QWEN3-4B2026.03 | 41.8 | |
| MoPPSModel=QWEN3-4B2026.03 | 41.59 | |
| INVERSE-EVIDENCEModel=QWEN3-4B2026.03 | 41.49 | |
| RANDOMModel=QWEN3-4B2026.03 | 41.13 | |
| Soft+Backbone=olmo-3-7b-it-sft2026.04 | 39.3 | |
| BaseBackbone=olmo-3.1-32b-it-sft2026.04 | 39.3 | |
| INSIGHTModel=R1-DISTILL-QWEN-7B2026.03 | 39.06 | |
| DSModel=R1-DISTILL-QWEN-7B2026.03 | 38.39 | |
| EXPECTED-DIFFICULTYModel=R1-DISTILL-QWEN-7B2026.03 | 38.39 | |
| Mag+Backbone=olmo-3-7b-it-sft2026.04 | 38.2 | |
| Pair+Backbone=olmo-3-7b-it-sft2026.04 | 38.2 | |
| MoPPSModel=R1-DISTILL-QWEN-7B2026.03 | 38.02 | |
| RANDOMModel=R1-DISTILL-QWEN-7B2026.03 | 37.98 | |
| INVERSE-EVIDENCEModel=R1-DISTILL-QWEN-7B2026.03 | 37.93 | |
| MPO+Backbone=olmo-3-7b-it-sft2026.04 | 34.9 | |
| DPOBackbone=olmo-3-7b-it-sft2026.04 | 34.6 | |
| RFTBackbone=olmo-3-7b-it-sft2026.04 | 33.1 | |
| BaseBackbone=olmo-3-7b-it-sft2026.04 | 30.5 | |
| INSIGHTModel=QWEN3-0.6B2026.03 | 21.94 | |
| INVERSE-EVIDENCEModel=QWEN3-0.6B2026.03 | 21.7 | |
| RANDOMModel=QWEN3-0.6B2026.03 | 21.6 | |
| MoPPSModel=QWEN3-0.6B2026.03 | 21.6 | |
| Soft+Backbone=gemma-3-4b-sft2026.04 | 21.3 | |
| Pair+Backbone=gemma-3-4b-sft2026.04 | 20.2 | |
| DPOBackbone=gemma-3-4b-sft2026.04 | 18.4 | |
| MPO+Backbone=gemma-3-4b-sft2026.04 | 18.4 | |
| BaseBackbone=gemma-3-4b-sft2026.04 | 17.6 | |
| Mag+Backbone=gemma-3-4b-sft2026.04 | 17.3 | |
| MARFTBackbone=LaMAS, Scale=Quad2025.04 | 16.99 | |
| RFTBackbone=gemma-3-4b-sft2026.04 | 16.9 | |
| MARFTBackbone=LaMAS, Scale=Trio2025.04 | 16.84 | |
| IPPO (MAPoRL)Backbone=LaMAS, Scale=Trio2025.04 | 16.76 | |
| IPPO (MAPoRL)Backbone=LaMAS, Scale=Duo2025.04 | 16.32 | |
| MARFTBackbone=LaMAS, Scale=Duo2025.04 | 16.25 | |
| VanillaBackbone=LaMAS, Scale=Trio2025.04 | 15.81 | |
| IPPO (MAPoRL)Backbone=LaMAS, Scale=Quad2025.04 | 15.81 | |
| VanillaBackbone=LaMAS, Scale=Quad2025.04 | 15.29 | |
| VanillaBackbone=LaMAS, Scale=Duo2025.04 | 13.46 |