Visual Attribution on ImageNet (val)
0.2185Deletion ScoreGrad-CAM
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Grad-CAMClassifier=ViT-B/32, Attribution Target=Ground Truth Class, Runtime=0.010s2026.04 | 0.2185 | 0.4931 | 0.2746 | 0.2647 | 0.5712 | 0.3065 | 63.3098 | 12.5057 | |
| ViT-CXClassifier=ViT-B/32, Attribution Target=Ground Truth Class, Runtime=0.763s2026.04 | 0.1924 | 0.5111 | 0.3187 | 0.2388 | 0.5806 | 0.3418 | 13.6714 | 45.4918 | |
| Trans-Att.Classifier=ViT-B/32, Attribution Target=Ground Truth Class, Runtime=0.157s2026.04 | 0.1886 | 0.5126 | 0.324 | 0.2306 | 0.5883 | 0.3576 | 15.2576 | 39.6312 | |
| MDAClassifier=ViT-B/32, Attribution Target=Ground Truth Class, Runtime=1.803s2026.04 | 0.1866 | 0.5443 | 0.3577 | 0.2298 | 0.6191 | 0.3893 | 24.5261 | 29.0981 | |
| LeGradClassifier=ViT-B/32, Attribution Target=Ground Truth Class, Runtime=0.005s2026.04 | 0.1792 | 0.5124 | 0.3331 | 0.2207 | 0.5876 | 0.3669 | 12.0416 | 43.7111 | |
| IGClassifier=ViT-B/32, Attribution Target=Ground Truth Class, Runtime=0.157s2026.04 | 0.1768 | 0.5157 | 0.3389 | 0.2181 | 0.591 | 0.3729 | 12.7448 | 43.2325 | |
| Bi-Att.Classifier=ViT-B/32, Attribution Target=Ground Truth Class, Runtime=0.158s2026.04 | 0.1762 | 0.5316 | 0.3554 | 0.2177 | 0.6071 | 0.3894 | 15.7758 | 38.8666 | |
| TISClassifier=ViT-B/32, Attribution Target=Ground Truth Class, Runtime=0.416s2026.04 | 0.1589 | 0.5691 | 0.4102 | 0.193 | 0.6535 | 0.4605 | 8.0883 | 53.2509 | |
| AHAClassifier=ViT-B/32, Attribution Target=Ground Truth Class, Runtime=0.016s2026.04 | 0.1363 | 0.525 | 0.3887 | 0.1694 | 0.6 | 0.4306 | 10.4213 | 43.8113 | |
| AHAClassifier=ViT-B/32, Attribution Target=Ground Truth Class, Refinement steps T=3, Runtime=0.212s2026.04 | 0.1171 | 0.565 | 0.448 | 0.142 | 0.6497 | 0.5077 | 7.008 | 51.5807 |