Image Classification on CIFAR-100 (Adversarial Robustness: Clean, PGD, C&W, AA)
66.12Clean AccuracyRAAT
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| RAATBackbone=ResNet-18, Threat Model=l22026.04 | 66.12 | 43.47 | 42.22 | 41.33 | 40.25 | |
| Cons-ATBackbone=ResNet-18, Threat Model=l22026.04 | 65.14 | 43.16 | 42.09 | 40.99 | 39.92 | |
| PGD-ATBackbone=ResNet-18, Threat Model=l22026.04 | 65 | 42.47 | 41.29 | 40.03 | 39.27 | |
| RAAT++Backbone=ResNet-18, Threat Model=l22026.04 | 62.43 | 44.95 | 43.96 | 42.35 | 40.61 | |
| TRADESBackbone=ResNet-18, Threat Model=l22026.04 | 61.25 | 43.52 | 42.96 | 41.22 | 40.15 | |
| MARTBackbone=ResNet-18, Threat Model=l22026.04 | 60.08 | 43.92 | 43.28 | 41.63 | 39.85 |