Medical on MedMCQA
58.2Accuracy (ACC)MAP
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| MAPMoE Architecture=DeepSeek-MoE, Category=Base2026.03 | 58.2 | 1.294 | 21.2 | 1.462 | |
| VGLR-FCMoE Architecture=DeepSeek-MoE, Category=Ours2026.03 | 56.2 | 1.146 | 3.8 | 0.124 | |
| VGLR-MFMoE Architecture=DeepSeek-MoE, Category=Ours2026.03 | 55.4 | 1.185 | 5.6 | 0.342 | |
| MAPMoE Architecture=Granite-MoE, Category=Base2026.03 | 55 | 1.29 | 18.3 | 0.288 | |
| MAPMoE Architecture=Qwen-MoE, Category=Base2026.03 | 54.2 | 1.285 | 19 | 0.286 | |
| Temp-ScaleMoE Architecture=DeepSeek-MoE, Category=Base2026.03 | 54.2 | 1.182 | 12.4 | 0.282 | |
| VTSRMoE Architecture=DeepSeek-MoE, Category=Ours2026.03 | 54 | 1.192 | 4.2 | 0.456 | |
| SWAGMoE Architecture=DeepSeek-MoE, Category=Weight2026.03 | 52.6 | 1.22 | 6.8 | 0.458 | |
| MCDRMoE Architecture=DeepSeek-MoE, Category=Weight2026.03 | 51.8 | 1.175 | 9.6 | 0.512 | |
| VGLR-MFMoE Architecture=Qwen-MoE, Category=Ours2026.03 | 49.7 | 1.178 | 9.7 | 0.145 | |
| MCDRMoE Architecture=Granite-MoE, Category=Weight2026.03 | 49.4 | 1.17 | 5 | 0.096 | |
| VGLR-FCMoE Architecture=Granite-MoE, Category=Ours2026.03 | 49.4 | 1.17 | 2.2 | 0.108 | |
| VGLR-MFMoE Architecture=Granite-MoE, Category=Ours2026.03 | 49.2 | 1.17 | 3.9 | 0.103 | |
| Temp-ScaleMoE Architecture=Qwen-MoE, Category=Base2026.03 | 49.1 | 1.179 | 12.2 | 0.193 | |
| SWAGMoE Architecture=Qwen-MoE, Category=Weight2026.03 | 49 | 1.212 | 18.8 | 0.28 | |
| VGLR-FCMoE Architecture=Qwen-MoE, Category=Ours2026.03 | 49 | 1.175 | 3 | 0.102 | |
| Temp-ScaleMoE Architecture=Granite-MoE, Category=Base2026.03 | 48.6 | 1.17 | 3.9 | 0.097 | |
| SWAGMoE Architecture=Granite-MoE, Category=Weight2026.03 | 48.6 | 1.2 | 9.6 | 0.179 | |
| MCDRMoE Architecture=Qwen-MoE, Category=Weight2026.03 | 48.2 | 1.174 | 12.5 | 0.097 | |
| VTSRMoE Architecture=Qwen-MoE, Category=Ours2026.03 | 47.9 | 1.167 | 5.1 | 0.103 | |
| VTSRMoE Architecture=Granite-MoE, Category=Ours2026.03 | 47.6 | 1.17 | 5.3 | 0.113 | |
| Full Informationk=Top-32026.06 | 0.886 | — | — | — | |
| Full Informationk=Top-22026.06 | 0.859 | — | — | — | |
| MeDxAgentk=Top-32026.06 | 0.8 | — | — | — | |
| MeDxAgentk=Top-22026.06 | 0.779 | — | — | — | |
| Full Informationk=Top-12026.06 | 0.763 | — | — | — | |
| DebateBackbone=Llama-3.3-70B-Instruct2025.05 | 0.75 | — | — | — | |
| CoTBackbone=Llama-3.3-70B-Instruct2025.05 | 0.738 | — | — | — | |
| SCBackbone=Llama-3.3-70B-Instruct2025.05 | 0.726 | — | — | — | |
| AgentVerseBackbone=Llama-3.3-70B-Instruct2025.05 | 0.726 | — | — | — | |
| DyLANBackbone=Llama-3.3-70B-Instruct2025.05 | 0.726 | — | — | — | |
| MAS-GPTBackbone=Llama-3.3-70B-Instruct2025.05 | 0.72 | — | — | — | |
| AgentVerseBackbone=Qwen-2.5-72B-Instruct2025.05 | 0.712 | — | — | — | |
| MADBackbone=Llama-3.3-70B-Instruct2025.05 | 0.71 | — | — | — | |
| DebateBackbone=Qwen-2.5-72B-Instruct2025.05 | 0.71 | — | — | — | |
| SingleBackbone=Llama-3.3-70B-Instruct2025.05 | 0.704 | — | — | — | |
| DyLANBackbone=Qwen-2.5-72B-Instruct2025.05 | 0.7 | — | — | — | |
| MeDxAgentk=Top-12026.06 | 0.698 | — | — | — | |
| MacNetBackbone=Llama-3.3-70B-Instruct2025.05 | 0.698 | — | — | — | |
| AutoGenBackbone=Llama-3.3-70B-Instruct2025.05 | 0.692 | — | — | — | |
| AFlow-MathBackbone=Llama-3.3-70B-Instruct2025.05 | 0.692 | — | — | — | |
| SCBackbone=Qwen-2.5-72B-Instruct2025.05 | 0.69 | — | — | — | |
| MAVBackbone=Llama-3.3-70B-Instruct2025.05 | 0.686 | — | — | — | |
| AutoGenBackbone=Qwen-2.5-72B-Instruct2025.05 | 0.684 | — | — | — | |
| CoTBackbone=Qwen-2.5-72B-Instruct2025.05 | 0.678 | — | — | — | |
| SingleBackbone=Qwen-2.5-72B-Instruct2025.05 | 0.676 | — | — | — | |
| MADBackbone=Qwen-2.5-72B-Instruct2025.05 | 0.674 | — | — | — | |
| AFlow-MathBackbone=Qwen-2.5-72B-Instruct2025.05 | 0.662 | — | — | — | |
| MAS-GPTBackbone=Qwen-2.5-72B-Instruct2025.05 | 0.662 | — | — | — | |
| MAVBackbone=Qwen-2.5-72B-Instruct2025.05 | 0.654 | — | — | — | |
| MacNetBackbone=Qwen-2.5-72B-Instruct2025.05 | 0.638 | — | — | — | |
| Best of FFT/LoRABase Model=Qwen-3B2026.05 | 0.6185 | — | — | — | |
| LoRA (best rank)Base Model=Qwen-3B2026.05 | 0.6185 | — | — | — | |
| NTK-SelectorModel=QWEN3-8B2025.11 | 0.614 | — | — | — | |
| Domain-OnlyModel=QWEN3-8B2025.11 | 0.611 | — | — | — | |
| DSIRModel=QWEN3-8B2025.11 | 0.611 | — | — | — | |
| EmbeddingModel=QWEN3-8B2025.11 | 0.609 | — | — | — | |
| RandomModel=QWEN3-8B2025.11 | 0.608 | — | — | — | |
| GradientModel=QWEN3-8B2025.11 | 0.607 | — | — | — | |
| MoLFBase Model=Qwen-3B2026.05 | 0.606 | — | — | — | |
| TSDSModel=QWEN3-8B2025.11 | 0.606 | — | — | — | |
| LESSModel=QWEN3-8B2025.11 | 0.603 | — | — | — | |
| FFTBase Model=Qwen-3B2026.05 | 0.5955 | — | — | — | |
| BaseModel=QWEN3-8B2025.11 | 0.593 | — | — | — | |
| NTK-SelectorModel=LLAMA3-8B-INSTRUCT2025.11 | 0.591 | — | — | — | |
| Best of FFT/LoRABase Model=Qwen-1.5B2026.05 | 0.5766 | — | — | — | |
| LoRA (best rank)Base Model=Qwen-1.5B2026.05 | 0.5766 | — | — | — | |
| DSIRModel=LLAMA3-8B-INSTRUCT2025.11 | 0.575 | — | — | — | |
| LESSModel=LLAMA3-8B-INSTRUCT2025.11 | 0.575 | — | — | — | |
| EmbeddingModel=LLAMA3-8B-INSTRUCT2025.11 | 0.574 | — | — | — | |
| GradientModel=LLAMA3-8B-INSTRUCT2025.11 | 0.574 | — | — | — | |
| RandomModel=LLAMA3-8B-INSTRUCT2025.11 | 0.571 | — | — | — | |
| TSDSModel=LLAMA3-8B-INSTRUCT2025.11 | 0.57 | — | — | — | |
| BaseModel=LLAMA3-8B-INSTRUCT2025.11 | 0.567 | — | — | — | |
| Domain-OnlyModel=LLAMA3-8B-INSTRUCT2025.11 | 0.565 | — | — | — | |
| MoLFBase Model=Qwen-1.5B2026.05 | 0.5628 | — | — | — | |
| FFTBase Model=Qwen-1.5B2026.05 | 0.5539 | — | — | — | |
| Best of FFT/LoRABase Model=Gemma-1B2026.05 | 0.4619 | — | — | — | |
| LoRA (best rank)Base Model=Gemma-1B2026.05 | 0.4619 | — | — | — | |
| MoLFBase Model=Gemma-1B2026.05 | 0.454 | — | — | — | |
| FFTBase Model=Gemma-1B2026.05 | 0.4215 | — | — | — |