Medical Image Classification on HAM10000
83.4AccuracyPS-CBM
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| PS-CBMBackbone=CLIP RN502025.11 | 83.4 | — | — | — | — | — | — | 61.3 | |
| DN-CBMBackbone=CLIP RN502025.11 | 80.5 | — | — | — | — | — | — | 47.6 | |
| Linear ProbeBackbone=CLIP RN502025.11 | 79.8 | — | — | — | — | — | — | — | |
| VLG-CBMBackbone=CLIP RN502025.11 | 79.8 | — | — | — | — | — | — | 59.9 | |
| Res-CBMBackbone=CLIP RN502025.11 | 77.5 | — | — | — | — | — | — | 52.7 | |
| LaBoBackbone=CLIP RN502025.11 | 77.1 | — | — | — | — | — | — | 50.7 | |
| LoRKDPre-train=Radimagenet, Params(M)=1.25, Comp. Ratio=5.32%2024.04 | 76.03 | — | — | — | — | — | — | — | |
| MTLPre-train=Radimagenet, Params(M)=1.25, Comp. Ratio=5.32%2024.04 | 75.83 | — | — | — | — | — | — | — | |
| FoundationPre-train=Med-MT, Params(M)=23.51, Comp. Ratio=/2024.04 | 75.53 | — | — | — | — | — | — | — | |
| V2C-CBMBackbone=CLIP RN502025.11 | 75.4 | — | — | — | — | — | — | 49.6 | |
| DCBMBackbone=CLIP RN502025.11 | 75.3 | — | — | — | — | — | — | 46.4 | |
| LoRKDPre-train=Med-MT, Params(M)=1.25, Comp. Ratio=5.32%2024.04 | 75.18 | — | — | — | — | — | — | — | |
| FoundationPre-train=Radimagenet, Params(M)=23.51, Comp. Ratio=/2024.04 | 75.08 | — | — | — | — | — | — | — | |
| MTLPre-train=Med-MT, Params(M)=1.25, Comp. Ratio=5.32%2024.04 | 74.92 | — | — | — | — | — | — | — | |
| MTLPre-train=MedMnist, Params(M)=1.25, Comp. Ratio=5.32%2024.04 | 74.82 | — | — | — | — | — | — | — | |
| LoRKDPre-train=MedMnist, Params(M)=1.25, Comp. Ratio=5.32%2024.04 | 74.82 | — | — | — | — | — | — | — | |
| Vanilla FS-ICLframework=Few-Shot In-Context Learning2025.06 | 74.76 | 23.7 | 70.51 | — | — | — | — | — | |
| CALINframework=Bi-level2025.06 | 74.76 | 2.68 | 74.24 | — | — | — | — | — | |
| BaselinePre-train=Radimagenet, Params(M)=1.25, Comp. Ratio=5.32%2024.04 | 74.42 | — | — | — | — | — | — | — | |
| BaselinePre-train=MedMnist, Params(M)=1.25, Comp. Ratio=5.32%2024.04 | 74.42 | — | — | — | — | — | — | — | |
| BaselinePre-train=Med-MT, Params(M)=1.25, Comp. Ratio=5.32%2024.04 | 74.42 | — | — | — | — | — | — | — | |
| STL-KDPre-train=Med-MT, Params(M)=1.25, Comp. Ratio=5.32%2024.04 | 74.42 | — | — | — | — | — | — | — | |
| MTL-KDPre-train=Radimagenet, Params(M)=1.25, Comp. Ratio=5.32%2024.04 | 74.37 | — | — | — | — | — | — | — | |
| MTL-KDPre-train=Med-MT, Params(M)=1.25, Comp. Ratio=5.32%2024.04 | 74.32 | — | — | — | — | — | — | — | |
| KFPre-train=Radimagenet, Params(M)=1.60, Comp. Ratio=6.81%2024.04 | 74.12 | — | — | — | — | — | — | — | |
| KFPre-train=Med-MT, Params(M)=1.60, Comp. Ratio=6.81%2024.04 | 73.92 | — | — | — | — | — | — | — | |
| STL-KDPre-train=MedMnist, Params(M)=1.25, Comp. Ratio=5.32%2024.04 | 73.87 | — | — | — | — | — | — | — | |
| MTL-KDPre-train=MedMnist, Params(M)=1.25, Comp. Ratio=5.32%2024.04 | 73.87 | — | — | — | — | — | — | — | |
| Aligned-MTLPre-train=MedMnist, Params(M)=1.25, Comp. Ratio=5.32%2024.04 | 73.87 | — | — | — | — | — | — | — | |
| Aligned-MTLPre-train=Med-MT, Params(M)=1.25, Comp. Ratio=5.32%2024.04 | 73.72 | — | — | — | — | — | — | — | |
| Aligned-MTLPre-train=Radimagenet, Params(M)=1.25, Comp. Ratio=5.32%2024.04 | 73.07 | — | — | — | — | — | — | — | |
| KFPre-train=MedMnist, Params(M)=1.60, Comp. Ratio=6.81%2024.04 | 72.97 | — | — | — | — | — | — | — | |
| MoCo-MTLPre-train=MedMnist, Params(M)=1.25, Comp. Ratio=5.32%2024.04 | 72.77 | — | — | — | — | — | — | — | |
| MoCo-MTLPre-train=Med-MT, Params(M)=1.25, Comp. Ratio=5.32%2024.04 | 72.62 | — | — | — | — | — | — | — | |
| FoundationPre-train=MedMnist, Params(M)=23.51, Comp. Ratio=/2024.04 | 72.12 | — | — | — | — | — | — | — | |
| STLPre-train=MedMnist, Params(M)=1.25, Comp. Ratio=5.32%2024.04 | 71.82 | — | — | — | — | — | — | — | |
| MoCo-MTLPre-train=Radimagenet, Params(M)=1.25, Comp. Ratio=5.32%2024.04 | 71.77 | — | — | — | — | — | — | — | |
| LF-CBMBackbone=CLIP RN502025.11 | 70.2 | — | — | — | — | — | — | 56.8 | |
| LM4CVBackbone=CLIP RN502025.11 | 66.8 | — | — | — | — | — | — | 49.8 | |
| FGG(best)Model=ResNet502024.03 | — | — | — | — | — | — | 66.14 | — | |
| FGG(best)Model=DeiT-B2024.03 | — | — | — | — | — | — | 63.93 | — | |
| GoGModel=ResNet502024.03 | — | — | — | — | — | — | 66.14 | — | |
| GoGModel=DeiT-B2024.03 | — | — | — | — | — | — | 63.93 | — | |
| GoUModel=ResNet502024.03 | — | — | — | — | — | — | 68.18 | — | |
| GoUModel=DeiT-B2024.03 | — | — | — | — | — | — | 64.95 | — | |
| Greedy SoupModel=ResNet502024.03 | — | — | — | — | — | — | 60.74 | — | |
| Greedy SoupModel=DeiT-B2024.03 | — | — | — | — | — | — | 64.87 | — | |
| GS(best)Model=ResNet502024.03 | — | — | — | — | — | — | 60.74 | — | |
| GS(best)Model=DeiT-B2024.03 | — | — | — | — | — | — | 64.87 | — | |
| Uniform SoupModel=ResNet502024.03 | — | — | — | — | — | — | 56.98 | — | |
| Uniform SoupModel=DeiT-B2024.03 | — | — | — | — | — | — | 14.29 | — |