Selective Classification on CIFAR-10 Clean original (test)
6.02AURCEV-AT
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| EV-ATBackbone=PreActResNet18, Data Augmentation=AugMix, Threat Model=l_inf2026.07 | 6.02 | 4.34 | 79.81 | |
| EV-ATBackbone=PreActResNet18, Data Augmentation=Cutout, Threat Model=l_inf2026.07 | 5.7 | 4.19 | 80.81 | |
| EV-ATBackbone=PreActResNet18, Data Augmentation=AutoAug, Threat Model=l_inf2026.07 | 5.18 | 3.85 | 80.03 | |
| EV-ATBackbone=PreActResNet18, Data Augmentation=Basic, Threat Model=l_inf2026.07 | 4.81 | 3.63 | 82.19 |