Adversarial Robustness and Fairness on CIFAR-100
75.17Clean Accuracy AvgNatural
Evaluation results
| Method | Links | |||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| NaturalBackbone=ResNet-182023.12 | 75.17 | 53.8 | 0.155 | 7.95 | 0 | 1.106 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | |
| BATBackbone=ResNet-182023.12 | 62.71 | 34.4 | 0.267 | 33.41 | 7.96 | 0.577 | 27.91 | 3.1 | 0.716 | 23.91 | 3 | 0.716 | 22.56 | 2.5 | 0.755 | |
| CFABackbone=ResNet-182023.12 | 59.29 | 33.8 | 0.245 | 33.88 | 10.3 | 0.531 | 31.15 | 8.4 | 0.563 | 26.85 | 5.2 | 0.655 | 25.83 | 4.9 | 0.679 | |
| AdaADBackbone=ResNet-182023.12 | 58.17 | 29.2 | 0.301 | 34.29 | 8.5 | 0.566 | 30.49 | 6.6 | 0.623 | 28.31 | 5 | 0.683 | 26.6 | 4.2 | 0.721 | |
| RSLADBackbone=ResNet-182023.12 | 57.88 | 29.4 | 0.302 | 34.5 | 9.2 | 0.55 | 31.19 | 7.4 | 0.598 | 28.13 | 4.7 | 0.669 | 26.46 | 3.7 | 0.703 | |
| Fair-ARDBackbone=ResNet-182023.12 | 57.19 | 29.4 | 0.303 | 33.57 | 8.4 | 0.558 | 30.57 | 7 | 0.598 | 27.96 | 4.5 | 0.666 | 26.15 | 3.9 | 0.713 | |
| SATBackbone=ResNet-182023.12 | 57.18 | 29.3 | 0.292 | 28.95 | 5.6 | 0.607 | 24.56 | 3.2 | 0.682 | 23.78 | 3.1 | 0.697 | 21.78 | 2.3 | 0.747 | |
| ABSLDBackbone=ResNet-182023.12 | 56.76 | 31.9 | 0.258 | 34.93 | 12.4 | 0.47 | 32.44 | 10.5 | 0.5 | 26.99 | 6.4 | 0.578 | 25.39 | 5.6 | 0.601 | |
| FRLBackbone=ResNet-182023.12 | 55.49 | 30.3 | 0.282 | 24.04 | 5 | 0.642 | 22.93 | 3.7 | 0.673 | 21.1 | 2.8 | 0.721 | — | — | — | |
| TRADESBackbone=ResNet-182023.12 | 55.33 | 28.4 | 0.303 | 30.5 | 7.3 | 0.559 | 27.71 | 5.6 | 0.655 | 24.33 | 3.5 | 0.736 | 23.55 | 3.2 | 0.756 |