Image Classification on CIFAR-10 L2 norm (epsilon=0.5, test)
95.74Accuracy (Standard)Rebuffi et al.
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Rebuffi et al.Backbone=WRN-70-16, Defense Type=Adversarial Training (AT), Extra data used=true2026.01 | 95.74 | 89.62 | 82.32 | |
| Gowal et al.Backbone=WRN-70-16, Defense Type=Adversarial Training (AT), Extra data used=true2026.01 | 94.74 | 88.18 | 80.53 | |
| Augustin et al.Backbone=WRN-28-10, Defense Type=Adversarial Training (AT)2026.01 | 93.96 | 86.14 | 78.79 | |
| Rebuffi et al.Backbone=WRN-70-16, Defense Type=Adversarial Training (AT)2026.01 | 92.41 | 86.24 | 80.42 | |
| Rebuffi et al.Backbone=WRN-28-10, Defense Type=Adversarial Training (AT), Extra data used=true2026.01 | 91.79 | 85.05 | 78.8 | |
| Sehwag et al.Backbone=WRN-28-10, Defense Type=Adversarial Training (AT)2026.01 | 90.93 | 83.75 | 77.24 |