Mathematical Reasoning on (GSM8K, SingleEq, SVAMP, MultiArith, AddSub, AQuA, MAWPS) (evaluation)
16.1GSM8K AccuracyBase Model
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Base ModelFT Dataset=Commonsense, Model=OLMoE-7B, Post-train Type=Base Model2026.04 | 16.1 | 23.6 | 17.7 | 9.2 | 21.3 | 22.8 | 13.9 | 17.8 | |
| ESFTFT Dataset=Commonsense, Model=OLMoE-7B, Post-train Type=ESFT2026.04 | 11.2 | 15.6 | 13.7 | 4.7 | 12.4 | 16.7 | 7.9 | 11.7 | |
| ExpertCondenserFT Dataset=Commonsense, Model=OLMoE-7B, Post-train Type=ExpertCondenser (Ours)2026.04 | 8.9 | 13.7 | 11.6 | 5.4 | 13.8 | 18.6 | 11.3 | 11.9 | |
| SFTFT Dataset=Commonsense, Model=OLMoE-7B, Post-train Type=SFT2026.04 | 6.8 | 14.6 | 12.5 | 5.7 | 13.4 | 17.8 | 8.6 | 11.3 | |
| DenseMixerFT Dataset=Commonsense, Model=OLMoE-7B, Post-train Type=DenseMixer2026.04 | 2.3 | 5.9 | 6.1 | 3.7 | 6.1 | 17.6 | 2.9 | 6.4 |