Image Classification on CIFAR-10 (test) (AutoAttack)
82.6AutoAttack AccuracyStochastic-LWTA
Evaluation results
| Method | Links | |
|---|---|---|
| Stochastic-LWTAtraining=PGD, backbone=WideResNet-34-102021.12 | 82.6 | |
| Stochastic-LWTAtraining=PGD, backbone=WideResNet-34-52021.12 | 81.22 | |
| Stochastic-LWTAtraining=PGD, backbone=WideResNet-34-12021.12 | 74.71 | |
| Gowal et al.additional unlabeled data=true2021.12 | 65.88 | |
| WARadditional unlabeled data=true2021.12 | 61.84 | |
| RSTadditional unlabeled data=true2021.12 | 59.53 | |
| HYDRAadditional unlabeled data=true2021.12 | 57.14 | |
| MARTadditional unlabeled data=true2021.12 | 56.29 | |
| Pre-trainingadditional unlabeled data=true2021.12 | 54.92 | |
| WARadditional unlabeled data=false2021.12 | 54.73 | |
| HE2021.12 | 53.74 | |
| FAT2021.12 | 53.51 | |
| Early-Stop2021.12 | 53.42 | |
| TRADES2021.12 | 53.08 |