Adversarial Robustness on CIFAR-100 (L-inf, epsilon=8/255, test)
73.85Clean AccuracyIKL-AT
Evaluation results
| Method | Links | ||
|---|---|---|---|
| IKL-ATArchitecture=WRN-28-10, Augmentation Type=50M Generated Data2023.05 | 73.85 | 39.18 | |
| DM-ATArchitecture=WRN-28-10, Augmentation Type=50M Generated Data2023.05 | 72.58 | 38.83 | |
| ACATArchitecture=WRN-34-10, Augmentation Type=AutoAug2023.05 | 68.74 | 31.3 | |
| IKL-ATArchitecture=WRN-34-10, Augmentation Type=AutoAug2023.05 | 66.08 | 32.53 | |
| IKL-ATArchitecture=WRN-34-10, Augmentation Type=Basic2023.05 | 65.76 | 31.91 | |
| ACATArchitecture=WRN-34-10, Augmentation Type=Basic2023.05 | 65.75 | 30.23 | |
| LAS-ATArchitecture=WRN-34-10, Augmentation Type=Basic2023.05 | 64.89 | 30.77 | |
| LBGATArchitecture=WRN-34-10, Augmentation Type=Basic2023.05 | 60.64 | 29.33 | |
| AWPArchitecture=WRN-34-10, Augmentation Type=Basic2023.05 | 60.38 | 28.86 |