Robust Classification on ImageNet-1K
66.2Accuracy (Natural)RoME+RANDOM
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| RoME+RANDOMBackbone=ViT-B, Attack Protocol=PGD2026.07 | 66.2 | 20.2 | 42.4 | 22 | 28.2 | 17 | |
| RANDOMBackbone=ViT-B, Attack Protocol=PGD2026.07 | 64.9 | 18.6 | 41.3 | 20.1 | 26.7 | 15.3 | |
| RoME+RANDOMBackbone=ViT-B, Attack Protocol=APGD2026.07 | 63.8 | 20.9 | 42.8 | 22.8 | 28.8 | 17.2 | |
| RANDOMBackbone=ViT-B, Attack Protocol=APGD2026.07 | 63.3 | 19.5 | 41.8 | 20.5 | 27.3 | 15.9 | |
| RoME+MAXBackbone=ViT-B, Attack Protocol=PGD2026.07 | 61.6 | 21.4 | 39.6 | 24.7 | 28.6 | 20.3 | |
| RoME+MAXBackbone=ViT-B, Attack Protocol=APGD2026.07 | 61.2 | 21.6 | 40.3 | 25.3 | 29.1 | 20.6 | |
| MAXBackbone=ViT-B, Attack Protocol=PGD2026.07 | 60.5 | 19.4 | 39.4 | 22.3 | 27 | 18.6 | |
| MAXBackbone=ViT-B, Attack Protocol=APGD2026.07 | 60.3 | 20 | 39.1 | 23.1 | 27.4 | 19 |