Adversarial Robustness on CIFAR-10 l_inf, epsilon=8/255 (test)
92.44Clean AccuracyDM-AT
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DM-ATArchitecture=WRN-28-10, Augmentation Type=20M Generated Data2025.03 | 92.44 | 67.31 | |
| GKL-ATArchitecture=WRN-28-10, Augmentation Type=20M Generated Data2025.03 | 92.16 | 67.75 | |
| LBGATArchitecture=WRN-34-20, Augmentation Type=Basic2025.03 | 88.7 | 53.57 | |
| ACATArchitecture=WRN-34-10, Augmentation Type=AutoAug2025.03 | 88.64 | 57.05 | |
| LAS-ATArchitecture=WRN-34-10, Augmentation Type=Basic2025.03 | 87.74 | 55.52 | |
| AWPArchitecture=WRN-34-10, Augmentation Type=Basic2025.03 | 85.36 | 56.17 | |
| Rice et al.Architecture=WRN-34-20, Augmentation Type=Basic2025.03 | 85.34 | 53.42 | |
| GKL-ATArchitecture=WRN-34-10, Augmentation Type=AutoAug2025.03 | 85.2 | 57.62 | |
| GKL-ATArchitecture=WRN-34-10, Augmentation Type=Basic2025.03 | 84.8 | 57.09 | |
| ACATArchitecture=WRN-34-10, Augmentation Type=Basic2025.03 | 82.41 | 55.36 |