Math on MATH (Score)
39.66ScoreDSR
Evaluation Results
| Method | Links | |
|---|---|---|
| DSRMidtraining Schedule=DSR2026.05 | 39.66 | |
| WSDMidtraining Schedule=WSD2026.05 | 36.86 | |
| Unified 15B MoEMidtraining Corpus=1k→8k2026.05 | 36.08 | |
| Cosine-decayMidtraining Schedule=Cos2026.05 | 33.98 | |
| Unified 15B MoEMidtraining Corpus=4k→8k2026.05 | 32.22 | |
| Unified 15B MoEMidtraining Corpus=4k2026.05 | 30.48 | |
| UM-190kBackbone=Apertus-8B-SFT, Evaluation Protocol=4-shot2025.11 | 11.59 | |
| UM-187kBackbone=Apertus-8B-SFT, Evaluation Protocol=4-shot2025.11 | 10.97 | |
| Unified 15B MoEPretraining Corpus=1k→8k2026.05 | 9.74 | |
| TuluDPOBackbone=Apertus-8B-SFT, Evaluation Protocol=4-shot2025.11 | 9.65 | |
| Unified 15B MoEPretraining Corpus=4k→8k2026.05 | 9.38 | |
| Unified 15B MoEPretraining Corpus=4k2026.05 | 9.2 | |
| UM-170kBackbone=Apertus-8B-SFT, Evaluation Protocol=4-shot2025.11 | 9.06 | |
| UltraFBBackbone=Apertus-8B-SFT, Evaluation Protocol=4-shot2025.11 | 8.7 | |
| ORPOBackbone=Apertus-8B-SFT, Evaluation Protocol=4-shot2025.11 | 8.06 | |
| CodePrefBackbone=Apertus-8B-SFT, Evaluation Protocol=4-shot2025.11 | 6.8 | |
| HelpSteerBackbone=Apertus-8B-SFT, Evaluation Protocol=4-shot2025.11 | 6.08 | |
| UM-190kshots=4-shot2025.11 | 6.05 | |
| ORPOshots=4-shot2025.11 | 5.87 | |
| UM-187kshots=4-shot2025.11 | 5.82 | |
| TuluDPOshots=4-shot2025.11 | 5.63 | |
| UltraFBshots=4-shot2025.11 | 5.44 | |
| UM-170kshots=4-shot2025.11 | 5.35 | |
| SFTBackbone=Apertus-8B-SFT, Evaluation Protocol=4-shot2025.11 | 4.53 | |
| SFTshots=4-shot2025.11 | 4.46 | |
| CodePrefshots=4-shot2025.11 | 4.42 | |
| HelpSteershots=4-shot2025.11 | 4.08 |