Image Classification on TAG → X average across unseen
56.2AccuracyOmniVaT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| OmniVaTBackbone=ViT-B / 16, Pre-trained=CLIP, Source Domain=TAG, Venue=Ours2026.01 | 56.2 | 53.7 | |
| PromptStylerBackbone=ViT-B / 16, Pre-trained=CLIP, Source Domain=TAG, Venue=ICCV-232026.01 | 51.7 | 40.6 | |
| LDCBackbone=ViT-B / 16, Pre-trained=CLIP, Source Domain=TAG, Venue=CVPR-252026.01 | 50.1 | 36.3 | |
| SPGBackbone=ViT-B / 16, Pre-trained=CLIP, Source Domain=TAG, Venue=ECCV-242026.01 | 49.4 | 41.9 | |
| CLIPBackbone=ViT-B / 16, Pre-trained=CLIP, Source Domain=TAG, Venue=ICML-212026.01 | 47.8 | 40 | |
| MMRLBackbone=ViT-B / 16, Pre-trained=CLIP, Source Domain=TAG, Venue=CVPR-252026.01 | 46 | 33.5 |