Attribution on ImageNet (val)
0.638Relevance RankSmooth Mask
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Smooth MaskModel=Supervised ResNet-502025.11 | 0.638 | 0.806 | 8.686 | 0.907 | 0.091 | — | — | |
| Smooth MaskModel=Supervised2025.11 | 0.638 | 0.806 | 8.686 | 0.907 | 0.091 | — | — | |
| Smooth MaskModel=DINO ViT-B/162025.11 | 0.616 | 0.758 | 8.659 | 0.91 | -0.052 | — | — | |
| Smooth MaskModel=DINO2025.11 | 0.616 | 0.758 | 8.659 | 0.91 | 0.01 | — | — | |
| Integrated GradsModel=Supervised ResNet-502025.11 | 0.614 | 0.632 | 10.163 | 0.589 | 0.07 | — | — | |
| Integrated GradsModel=Supervised2025.11 | 0.614 | 0.632 | 10.163 | 0.589 | 0.07 | — | — | |
| Gradient SHAPModel=DINO ViT-B/162025.11 | 0.61 | 0.633 | 10.152 | 0.589 | 0.005 | — | — | |
| Integrated GradsModel=DINO ViT-B/162025.11 | 0.61 | 0.637 | 10.21 | 0.571 | -0.016 | — | — | |
| Gradient SHAPModel=DINO2025.11 | 0.61 | 0.633 | 10.152 | 0.589 | -0.05 | — | — | |
| Integrated GradsModel=DINO2025.11 | 0.61 | 0.637 | 10.21 | 0.571 | 0 | — | — | |
| Gradient SHAPModel=Supervised ResNet-502025.11 | 0.606 | 0.628 | 10.115 | 0.604 | 0.036 | — | — | |
| Gradient SHAPModel=Supervised2025.11 | 0.606 | 0.628 | 10.115 | 0.604 | 0.036 | — | — | |
| Extremal ContourModel=DINO ViT-B/162025.11 | 0.601 | 0.814 | 8.562 | 0.899 | 0.045 | — | — | |
| Extremal ContourModel=DINO2025.11 | 0.601 | 0.814 | 8.562 | 0.899 | -0.033 | — | — | |
| Extremal ContourModel=Supervised ResNet-502025.11 | 0.596 | 0.779 | 8.878 | 0.855 | 0.05 | — | — | |
| Extremal ContourModel=Supervised2025.11 | 0.596 | 0.779 | 8.878 | 0.855 | 0.05 | — | — | |
| Grad-CAM++Model=Supervised ResNet-502025.11 | 0.588 | 0.626 | 9.445 | 0.658 | 0.09 | — | — | |
| Grad-CAM++Model=Supervised2025.11 | 0.588 | 0.626 | 9.737 | 0.658 | 0.09 | — | — | |
| Grad-CAM++Model=DINO ViT-B/162025.11 | 0.551 | 0.584 | 10.022 | 0.636 | 0.014 | — | — | |
| Grad-CAM++Model=DINO2025.11 | 0.551 | 0.584 | 10.022 | 0.636 | -0.041 | — | — | |
| HSIC-AttributionBackbone=ViT-L [74]2025.04 | — | — | — | — | — | 0.1898 | 0.5253 | |
| HSIC-AttributionBackbone=Swin-L [75]2025.04 | — | — | — | — | — | 0.3589 | 0.5531 | |
| Integrated GradientsBackbone=ViT-L [74]2025.04 | — | — | — | — | — | 0.153 | 0.2402 | |
| Integrated GradientsBackbone=Swin-L [75]2025.04 | — | — | — | — | — | 0.3074 | 0.4198 | |
| Kernel SHAPBackbone=ViT-L [74]2025.04 | — | — | — | — | — | 0.4155 | 0.4942 | |
| Kernel SHAPBackbone=Swin-L [75]2025.04 | — | — | — | — | — | 0.43 | 0.4794 | |
| LIMABackbone=ViT-L [74], variant=w/ SLICO2025.04 | — | — | — | — | — | 0.0994 | 0.7405 | |
| LIMABackbone=Swin-L [75], variant=w/ SLICO2025.04 | — | — | — | — | — | 0.1371 | 0.7627 | |
| Previous workBackbone=ViT-L [74]2025.04 | — | — | — | — | — | 0.1268 | 0.6715 | |
| Previous workBackbone=Swin-L [75]2025.04 | — | — | — | — | — | 0.249 | 0.6811 | |
| RISEBackbone=ViT-L [74]2025.04 | — | — | — | — | — | 0.1947 | 0.6177 | |
| RISEBackbone=Swin-L [75]2025.04 | — | — | — | — | — | 0.3338 | 0.5655 | |
| SaliencyBackbone=ViT-L [74]2025.04 | — | — | — | — | — | 0.3499 | 0.3544 | |
| SaliencyBackbone=Swin-L [75]2025.04 | — | — | — | — | — | 0.2888 | 0.4178 |