Chest X-ray Classification on Tuberculosis (test)
99.7AccuracyLKA
Evaluation results
| Method | Links | ||
|---|---|---|---|
| LKABackbone=Swin-L, #Trainable Parameters=0.652M2025.06 | 99.7 | — | |
| Full fine-tuningBackbone=Swin-L, #Trainable Parameters=195M2025.06 | 99.5 | — | |
| BitfitBackbone=Swin-L, #Trainable Parameters=0.313M2025.06 | 99.3 | — | |
| LORABackbone=Swin-L, #Trainable Parameters=0.578M2025.06 | 99.3 | — | |
| ConvpassBackbone=Swin-L, #Trainable Parameters=0.661M2025.06 | 99.3 | — | |
| AdapterformerBackbone=Swin-L, #Trainable Parameters=0.320M2025.06 | 99.1 | — | |
| LOSSLESS ADAPTATIONBackbone=Swin-L, #Trainable Parameters=0.320M2025.06 | 99.1 | — | |
| CIATBackbone=Swin-L, #Trainable Parameters=0.966M2025.06 | 99 | — | |
| RepAdapterBackbone=Swin-L, #Trainable Parameters=0.486M2025.06 | 98.8 | — | |
| AdapterBackbone=Swin-L, #Trainable Parameters=0.633M2025.06 | 98.7 | — | |
| VPTBackbone=Swin-L, #Trainable Parameters=1.052M2025.06 | 98.7 | — | |
| AIMBackbone=Swin-L, #Trainable Parameters=0.947M2025.06 | 97.2 | — | |
| Linear probingBackbone=Swin-L, #Trainable Parameters=0.006M2025.06 | 93.9 | — | |
| ST-AdapterBackbone=Swin-L, #Trainable Parameters=0.334M2025.06 | 93.3 | — | |
| LoRAEvaluation Protocol=Fine-tuned individually2025.03 | 81.8 | 91.6 | |
| SVD-LoRAEvaluation Protocol=Fine-tuned individually2025.03 | 81.8 | 91.5 | |
| ViTEvaluation Protocol=Fine-tuned individually2025.03 | 81.2 | 89.4 | |
| Med-LEGOEvaluation Protocol=Single merged model2025.03 | 80.3 | 89.7 | |
| PEM CompositionEvaluation Protocol=Single merged model2025.03 | 77.3 | 80.6 | |
| MagMaxEvaluation Protocol=Single merged model2025.03 | 71.2 | 78.3 | |
| Task ArithmeticEvaluation Protocol=Single merged model2025.03 | 53 | 52.7 | |
| ViT AveragingEvaluation Protocol=Single merged model2025.03 | 51.5 | 54.2 | |
| Ties-MergingEvaluation Protocol=Single merged model2025.03 | 50 | 47.2 |