Common Sense on OpenBookQA
81.8AccuracyTemp-Scale
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Temp-ScaleMoE Architecture=Qwen-MoE, Category=Base2026.03 | 81.8 | 1.12 | 10.2 | 22.4 | |
| VTSRMoE Architecture=Qwen-MoE, Category=Ours2026.03 | 81.8 | 0.682 | 2.2 | 10.6 | |
| Temp-ScaleMoE Architecture=DeepSeek-MoE, Category=Base2026.03 | 80.8 | 0.87 | 14.6 | 40.2 | |
| MAPMoE Architecture=Qwen-MoE, Category=Base2026.03 | 80.4 | 1.37 | 12.7 | 27.3 | |
| VGLR-FCMoE Architecture=Qwen-MoE, Category=Ours2026.03 | 80.2 | 0.694 | 1.4 | 9.8 | |
| MAPMoE Architecture=DeepSeek-MoE, Category=Base2026.03 | 80.2 | 1.12 | 16.8 | 32.1 | |
| VGLR-FCMoE Architecture=DeepSeek-MoE, Category=Ours2026.03 | 80 | 0.447 | 5.4 | 14.6 | |
| MCDRMoE Architecture=Qwen-MoE, Category=Weight2026.03 | 79.8 | 0.982 | 6.4 | 15.2 | |
| VTSRMoE Architecture=DeepSeek-MoE, Category=Ours2026.03 | 79.8 | 0.431 | 6 | 26.2 | |
| VGLR-MFMoE Architecture=Qwen-MoE, Category=Ours2026.03 | 79.6 | 0.724 | 2.8 | 12.4 | |
| MCDRMoE Architecture=DeepSeek-MoE, Category=Weight2026.03 | 79.4 | 0.729 | 10.3 | 30.2 | |
| VGLR-MFMoE Architecture=DeepSeek-MoE, Category=Ours2026.03 | 78.8 | 0.476 | 6.7 | 17.4 | |
| SWAGMoE Architecture=DeepSeek-MoE, Category=Weight2026.03 | 78.6 | 0.665 | 14 | 28.8 | |
| SWAGMoE Architecture=Qwen-MoE, Category=Weight2026.03 | 78.2 | 0.912 | 9.8 | 14.6 | |
| MAPMoE Architecture=Granite-MoE, Category=Base2026.03 | 74.6 | 1.38 | 25.2 | 47.2 | |
| VGLR-MFMoE Architecture=Granite-MoE, Category=Ours2026.03 | 74.2 | 0.654 | 2.6 | 29.3 | |
| VGLR-FCMoE Architecture=Granite-MoE, Category=Ours2026.03 | 74 | 0.652 | 1.5 | 15.2 | |
| SWAGMoE Architecture=Granite-MoE, Category=Weight2026.03 | 73.6 | 0.652 | 4.1 | 29 | |
| VTSRMoE Architecture=Granite-MoE, Category=Ours2026.03 | 73.6 | 0.667 | 5.2 | 29.3 | |
| MCDRMoE Architecture=Granite-MoE, Category=Weight2026.03 | 73.4 | 0.65 | 3.7 | 29.8 | |
| Temp-ScaleMoE Architecture=Granite-MoE, Category=Base2026.03 | 71.6 | 0.773 | 10.7 | 20.1 |