Feature Importance Assessment on ImageNet-A (test)
9.67InsertionViT-CX
Evaluation results
| Method | Links | ||
|---|---|---|---|
| ViT-CXModel=ViT†-B-16/224, Training Scheme=CLIP2025.07 | 9.67 | 10.98 | |
| Chefer-LRPModel=ViT*-L-16/224, Training Scheme=CLIP + Massive Regularization2025.07 | 3.99 | 32.74 | |
| CoIBAModel=EVA-L-14/196, Training Scheme=EVA2025.07 | 3.65 | 50.41 | |
| CoIBAModel=ViT*-L-16/224, Training Scheme=CLIP + Massive Regularization2025.07 | 2.53 | 41.07 | |
| Chefer-LRPModel=ViT*-B-16/224, Training Scheme=CLIP + Massive Regularization2025.07 | 2.46 | 25.99 | |
| CoIBAModel=ViT†-B-16/224, Training Scheme=CLIP2025.07 | 1.97 | 36.15 | |
| CoIBAModel=ViT*-B-16/224, Training Scheme=CLIP + Massive Regularization2025.07 | 1.82 | 32.62 |