Few-shot Image Classification on CD-FSL 5-way 5-shot (test)
28.93ChestX AccuracyIM-DCL
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| IM-DCLbackbone=RN10, Source=-, FT=Y2026.03 | 28.93 | 52.74 | 89.47 | 95.73 | 66.72 | |
| AttnTempbackbone=ViT/DINO, Source=Y, FT=Y2026.03 | 28.03 | 54.91 | 90.82 | 96.66 | 67.61 | |
| CD-CLSbackbone=ViT/DINO, Source=Y, FT=Y2026.03 | 27.66 | 54.69 | 91.53 | 96.27 | 67.54 | |
| DARAbackbone=RN10, Source=Y, FT=Y2026.03 | 27.54 | 56.28 | 85.84 | 95.32 | 66.25 | |
| PMFbackbone=ViT/DINO, Source=Y, FT=Y2026.03 | 27.27 | 50.12 | 85.98 | 92.96 | 64.08 | |
| PE-Core-LoRA + OURSbackbone=ViT/PE-Core, Source=-, FT=Y2026.03 | 27.05 | 60.03 | 94.67 | 98.36 | 70.05 | |
| FLoRbackbone=ViT/DINO, Source=Y, FT=Y2026.03 | 27.02 | 53.06 | 90.75 | 96.47 | 66.83 | |
| StyleAdv-FTbackbone=ViT/DINO, Source=Y, FT=Y2026.03 | 26.97 | 51.23 | 90.12 | 95.99 | 66.08 | |
| CLIP-LoRA-Vision + OURSbackbone=ViT/CLIP, Source=-, FT=Y2026.03 | 26.42 | 56.2 | 94.58 | 97.21 | 68.57 | |
| StepSTPbackbone=ViT/CLIP, Source=-, FT=Y2026.03 | 26.36 | 52.12 | 89.4 | 96.01 | 65.97 | |
| StyleAdv-FTbackbone=RN10, Source=Y, FT=Y2026.03 | 26.24 | 53.05 | 91.64 | 96.51 | 66.86 | |
| StyleAdvbackbone=RN10, Source=Y, FT=-2026.03 | 26.07 | 45.77 | 86.58 | 93.65 | 63.02 | |
| LDCbackbone=ViT/CLIP, Source=-, FT=Y2026.03 | 25.89 | 49.7 | 90.82 | 96.71 | 65.78 | |
| wave-SANbackbone=RN10, Source=Y, FT=-2026.03 | 25.63 | 44.93 | 85.22 | 89.7 | 61.37 | |
| FN+VDBbackbone=RN18, Source=-, FT=Y2026.03 | 25.55 | 47.48 | 87.31 | 94.63 | 64.74 | |
| SigLIP2-LoRA + OURSbackbone=ViT/SigLip2, Source=-, FT=Y2026.03 | 25.54 | 55.11 | 92.7 | 97.63 | 67.75 | |
| ATA-FTbackbone=RN10, Source=Y, FT=Y2026.03 | 25.08 | 49.79 | 89.64 | 95.44 | 64.99 | |
| AFAbackbone=RN10, Source=Y, FT=-2026.03 | 25.02 | 46.01 | 85.58 | 88.06 | 61.17 | |
| CLIP-LoRA-Visionbackbone=ViT/CLIP, Source=-, FT=Y2026.03 | 24.61 | 52.22 | 93.31 | 95.88 | 66.5 | |
| PE-Core-LoRAbackbone=ViT/PE-Core, Source=-, FT=Y2026.03 | 24.44 | 58.81 | 94.07 | 97.25 | 68.64 | |
| SEALParameters Updated=Linear, Backbone=ResNet-50, Pre-training Dataset=Tiny-ImageNet, Episodes=6002023.04 | 24.42 | 37.58 | 74.61 | 85 | — | |
| ATAbackbone=RN10, Source=Y, FT=-2026.03 | 24.32 | 44.91 | 83.75 | 90.59 | 60.89 | |
| SigLIP2-LoRAbackbone=ViT/SigLip2, Source=-, FT=Y2026.03 | 24.24 | 51.79 | 91.39 | 96.43 | 65.96 | |
| Tip-Adapterbackbone=ViT/CLIP, Source=-, FT=Y2026.03 | 24.07 | 46.96 | 87.24 | 94.19 | 63.12 | |
| Maple + OURSbackbone=ViT/CLIP, Source=-, FT=Y2026.03 | 24.04 | 49.81 | 92.24 | 94.62 | 65.18 | |
| LP++backbone=ViT/CLIP, Source=-, FT=Y2026.03 | 23.89 | 48.49 | 87.48 | 94.47 | 63.58 | |
| PromptSRCbackbone=ViT/CLIP, Source=-, FT=Y2026.03 | 23.51 | 46.09 | 86.54 | 89.97 | 61.52 | |
| CoOpbackbone=ViT/CLIP, Source=-, FT=Y2026.03 | 23.35 | 45.78 | 85.88 | 93.31 | 62.08 | |
| AMU-Tuningbackbone=ViT/CLIP, Source=-, FT=Y2026.03 | 23.34 | 44.6 | 88.47 | 94.26 | 62.66 | |
| SEALParameters Updated=Linear+Affines, Backbone=ResNet-50, Pre-training Dataset=Tiny-ImageNet, Episodes=6002023.04 | 22.98 | 39.64 | 73.83 | 88.24 | — | |
| Maplebackbone=ViT/CLIP, Source=-, FT=Y2026.03 | 22.96 | 48.35 | 89.04 | 93.5 | 63.46 | |
| NormalParameters Updated=Linear, Backbone=ResNet-50, Pre-training Dataset=Tiny-ImageNet, Episodes=6002023.04 | 22.79 | 33.28 | 71.74 | 77.05 | — | |
| LLFParameters Updated=Linear, Backbone=ResNet-50, Pre-training Dataset=Tiny-ImageNet, Episodes=6002023.04 | 22.03 | 29.6 | 60.69 | 53.55 | — | |
| PDAbackbone=ViT/CLIP, Source=-, FT=Y2026.03 | 21.87 | 45.19 | 86.21 | 92.67 | 61.48 | |
| Normal (long)Parameters Updated=Linear+Affines, Backbone=ResNet-50, Pre-training Dataset=Tiny-ImageNet, Episodes=6002023.04 | 21.59 | 29.67 | 55.79 | 71.55 | — | |
| Normal (long)Parameters Updated=Linear, Backbone=ResNet-50, Pre-training Dataset=Tiny-ImageNet, Episodes=6002023.04 | 21 | 28.93 | 53.4 | 57.1 | — | |
| NormalParameters Updated=Linear+Affines, Backbone=ResNet-50, Pre-training Dataset=Tiny-ImageNet, Episodes=6002023.04 | 20.82 | 28.57 | 53.35 | 63.22 | — | |
| LLFParameters Updated=Linear+Affines, Backbone=ResNet-50, Pre-training Dataset=Tiny-ImageNet, Episodes=6002023.04 | 20.44 | 22.01 | 30.83 | 28.06 | — |