Image Classification on CIFAR-10-C
93.32AccuracyVanilla Train
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Vanilla TrainAug.=L2, Pruning Ratio=50%2026.01 | 93.32 | — | — | — | — | — | — | 86 | — | — | |
| Vanilla TrainAug.=L2, Pruning Ratio=70%2026.01 | 93.32 | — | — | — | — | — | — | 86 | — | — | |
| Vanilla TrainAug.=None, Pruning Ratio=50%2026.01 | 93.04 | — | — | — | — | — | — | 26 | — | — | |
| Vanilla TrainAug.=None, Pruning Ratio=70%2026.01 | 93.04 | — | — | — | — | — | — | 26 | — | — | |
| C-SAMAug.=L2, Pruning Ratio=50%2026.01 | 92.7 | — | — | — | — | — | — | 82 | — | — | |
| Vanilla TrainAug.=L1, Pruning Ratio=50%2026.01 | 91.84 | — | — | — | — | — | — | 80 | — | — | |
| Vanilla TrainAug.=L1, Pruning Ratio=70%2026.01 | 91.84 | — | — | — | — | — | — | 80 | — | — | |
| C-SAMAug.=L2, Pruning Ratio=70%2026.01 | 91.77 | — | — | — | — | — | — | 67 | — | — | |
| defensive patch generation frameworkModel=VGG, Setting=Four models ensemble2022.04 | 91.02 | — | — | — | — | — | — | — | — | — | |
| C-SAMAug.=L1, Pruning Ratio=70%2026.01 | 89.42 | — | — | — | — | — | — | 57 | — | — | |
| C-SAMAug.=L1, Pruning Ratio=50%2026.01 | 88.16 | — | — | — | — | — | — | 60 | — | — | |
| LMPAug.=None, Pruning Ratio=50%2026.01 | 87.53 | — | — | — | — | — | — | 9 | — | — | |
| defensive patch generation frameworkModel=MNet, Setting=Four models ensemble2022.04 | 87.37 | — | — | — | — | — | — | — | — | — | |
| AdaBNTraining Augmentations=true2021.03 | 86.7 | — | — | — | — | — | — | — | — | — | |
| LMPAug.=None, Pruning Ratio=70%2026.01 | 86.65 | — | — | — | — | — | — | 6 | — | — | |
| LMPAug.=L2, Pruning Ratio=50%2026.01 | 86.63 | — | — | — | — | — | — | 56 | — | — | |
| LMPAug.=L1, Pruning Ratio=50%2026.01 | 86.47 | — | — | — | — | — | — | 51 | — | — | |
| HYDRAAug.=L2, Pruning Ratio=50%2026.01 | 86.43 | — | — | — | — | — | — | 64 | — | — | |
| HYDRAAug.=None, Pruning Ratio=50%2026.01 | 86.02 | — | — | — | — | — | — | 35 | — | — | |
| LMPAug.=L1, Pruning Ratio=70%2026.01 | 86.02 | — | — | — | — | — | — | 48 | — | — | |
| HYDRAAug.=L1, Pruning Ratio=50%2026.01 | 85.99 | — | — | — | — | — | — | 53 | — | — | |
| HYDRAAug.=None, Pruning Ratio=70%2026.01 | 85.47 | — | — | — | — | — | — | 29 | — | — | |
| LMPAug.=L2, Pruning Ratio=70%2026.01 | 84.64 | — | — | — | — | — | — | 46 | — | — | |
| HYDRAAug.=L2, Pruning Ratio=70%2026.01 | 84.52 | — | — | — | — | — | — | 49 | — | — | |
| HYDRAAug.=L1, Pruning Ratio=70%2026.01 | 84.41 | — | — | — | — | — | — | 50 | — | — | |
| AdaBNTraining Augmentations=false2021.03 | 83.6 | — | — | — | — | — | — | — | — | — | |
| defensive patch generation frameworkModel=SNet, Setting=Four models ensemble2022.04 | 83.26 | — | — | — | — | — | — | — | — | — | |
| SurgeonBackbone=ViT-Base, Evaluation Protocol=Test-Time Adaptation2025.10 | 82.7 | — | — | — | — | — | — | — | — | — | |
| NEOBackbone=ViT-Base, Evaluation Protocol=Test-Time Adaptation2025.10 | 82.4 | — | — | — | — | — | — | — | — | — | |
| TENTBackbone=ViT-Base, Evaluation Protocol=Test-Time Adaptation2025.10 | 81.3 | — | — | — | — | — | — | — | — | — | |
| FOABackbone=ViT-Base, Evaluation Protocol=Test-Time Adaptation2025.10 | 80.9 | — | — | — | — | — | — | — | — | — | |
| S2-SAMAug.=None, Pruning Ratio=50%2026.01 | 80.63 | — | — | — | — | — | — | 4 | — | — | |
| SARBackbone=ViT-Base, Evaluation Protocol=Test-Time Adaptation2025.10 | 80.6 | — | — | — | — | — | — | — | — | — | |
| CoTTABackbone=ViT-Base, Evaluation Protocol=Test-Time Adaptation2025.10 | 80.6 | — | — | — | — | — | — | — | — | — | |
| No AdaptBackbone=ViT-Base, Evaluation Protocol=Test-Time Adaptation2025.10 | 80.4 | — | — | — | — | — | — | — | — | — | |
| T3ABackbone=ViT-Base, Evaluation Protocol=Test-Time Adaptation2025.10 | 80.1 | — | — | — | — | — | — | — | — | — | |
| WATT-SBackbone=ViT-L/142024.06 | 80.06 | — | — | — | — | — | — | — | — | — | |
| WATT-PBackbone=ViT-L/142024.06 | 80.05 | — | — | — | — | — | — | — | — | — | |
| S2-SAMAug.=L1, Pruning Ratio=50%2026.01 | 79.88 | — | — | — | — | — | — | 33 | — | — | |
| LAMEBackbone=ViT-Base, Evaluation Protocol=Test-Time Adaptation2025.10 | 79.8 | — | — | — | — | — | — | — | — | — | |
| S2-SAMAug.=L2, Pruning Ratio=50%2026.01 | 79.42 | — | — | — | — | — | — | 21 | — | — | |
| TENTBackbone=ViT-L/142024.06 | 79.18 | — | — | — | — | — | — | — | — | — | |
| Ours2024.10 | 78.33 | — | — | — | — | — | — | — | — | — | |
| S2-SAMAug.=L1, Pruning Ratio=70%2026.01 | 78.26 | — | — | — | — | — | — | 26 | — | — | |
| S2-SAMAug.=None, Pruning Ratio=70%2026.01 | 78.21 | — | — | — | — | — | — | 4 | — | — | |
| CLIPArTTBackbone=ViT-L/142024.06 | 78.06 | — | — | — | — | — | — | — | — | — | |
| FFM2024.10 | 77.77 | — | — | — | — | — | — | — | — | — | |
| TRUST2025.09 | 77.5 | — | — | — | — | — | — | — | — | — | |
| SADA2024.10 | 77.33 | — | — | — | — | — | — | — | — | — | |
| S2-SAMAug.=L2, Pruning Ratio=70%2026.01 | 77.09 | — | — | — | — | — | — | 28 | — | — | |
| WATT-SBackbone=ViT-B/162024.06 | 76.22 | — | — | — | — | — | — | — | — | — | |
| defensive patch generation frameworkModel=RNet, Setting=Four models ensemble2022.04 | 76.04 | — | — | — | — | — | — | — | — | — | |
| CLIPBackbone=ViT-L/142024.06 | 76.04 | — | — | — | — | — | — | — | — | — | |
| AugMix2024.10 | 75.28 | — | — | — | — | — | — | — | — | — | |
| AdvTrain+LMPAug.=None, Pruning Ratio=50%2026.01 | 75.13 | — | — | — | — | — | — | 0 | — | — | |
| AdvTrainAug.=L2, Pruning Ratio=50%2026.01 | 75.05 | — | — | — | — | — | — | 1 | — | — | |
| AdvTrainAug.=L2, Pruning Ratio=70%2026.01 | 75.05 | — | — | — | — | — | — | 1 | — | — | |
| WATT-PBackbone=ViT-B/162024.06 | 75.04 | — | — | — | — | — | — | — | — | — | |
| TPTBackbone=ViT-L/14, Batch Size=322024.06 | 75.01 | — | — | — | — | — | — | — | — | — | |
| Δ-UQBackbone=ResNet-202022.07 | 74.25 | — | — | — | — | — | — | — | — | — | |
| TRUST naive2025.09 | 74.2 | — | — | — | — | — | — | — | — | — | |
| AdvTrain+LMPAug.=L1, Pruning Ratio=50%2026.01 | 74.17 | — | — | — | — | — | — | 12 | — | — | |
| AdvTrain+LMPAug.=L2, Pruning Ratio=70%2026.01 | 74.06 | — | — | — | — | — | — | 3 | — | — | |
| WATT-SMechanism=Sequential MTWA2024.06 | 73.82 | — | — | — | — | — | — | — | — | — | |
| AdvTrainAug.=L1, Pruning Ratio=50%2026.01 | 73.37 | — | — | — | — | — | — | 2 | — | — | |
| AdvTrainAug.=L1, Pruning Ratio=70%2026.01 | 73.37 | — | — | — | — | — | — | 2 | — | — | |
| CLIPArTTBackbone=ViT-B/162024.06 | 73.22 | — | — | — | — | — | — | — | — | — | |
| AdvTrainAug.=None, Pruning Ratio=50%2026.01 | 73.09 | — | — | — | — | — | — | 0 | — | — | |
| AdvTrainAug.=None, Pruning Ratio=70%2026.01 | 73.09 | — | — | — | — | — | — | 0 | — | — | |
| DEnsBackbone=ResNet-202022.07 | 72.9 | — | — | — | — | — | — | — | — | — | |
| L2D2024.10 | 72.88 | — | — | — | — | — | — | — | — | — | |
| WATT-PMechanism=Parallel MTWA2024.06 | 72.83 | — | — | — | — | — | — | — | — | — | |
| AdvTrain+LMPAug.=L2, Pruning Ratio=50%2026.01 | 72.23 | — | — | — | — | — | — | 3 | — | — | |
| FedSelectBackbone=ResNet-18, Initialization=Random2024.04 | 72.05 | — | — | — | — | — | — | — | — | — | |
| AdvTrain+LMPAug.=None, Pruning Ratio=70%2026.01 | 72.02 | — | — | — | — | — | — | 0 | — | — | |
| Flying Bird+Aug.=L2, Pruning Ratio=70%2026.01 | 71.77 | — | — | — | — | — | — | 34 | — | — | |
| BatchEnsemblesBackbone=WRN-28-10, Cutout=true2020.12 | 71.67 | 0.1928 | — | — | — | — | — | — | — | — | |
| RandConv2024.10 | 71.23 | — | — | — | — | — | — | — | — | — | |
| CLIPARTT2024.06 | 71.17 | — | — | — | — | — | — | — | — | — | |
| AdvTrain+LMPAug.=L1, Pruning Ratio=70%2026.01 | 70.9 | — | — | — | — | — | — | 2 | — | — | |
| SVIBackbone=ResNet-202022.07 | 70.6 | — | — | — | — | — | — | — | — | — | |
| UnAdvModel=VGG, Setting=Four models ensemble2022.04 | 70.36 | — | — | — | — | — | — | — | — | — | |
| MC DropoutBackbone=ResNet-202022.07 | 70.2 | — | — | — | — | — | — | — | — | — | |
| Flying Bird+Aug.=L2, Pruning Ratio=50%2026.01 | 70.06 | — | — | — | — | — | — | 30 | — | — | |
| MIMOBackbone=WRN-28-10, Cutout=false2020.12 | 69.99 | 0.1846 | — | — | — | — | — | — | — | — | |
| VanillaBackbone=ResNet-202022.07 | 69.8 | — | — | — | — | — | — | — | — | — | |
| LP-BNNBackbone=WRN-28-10, Cutout=true2020.12 | 69.51 | 0.1197 | — | — | — | — | — | — | — | — | |
| Local OnlyBackbone=ResNet-18, Initialization=Random2024.04 | 69.5 | — | — | — | — | — | — | — | — | — | |
| FedPACBackbone=ResNet-18, Initialization=Random2024.04 | 69.5 | — | — | — | — | — | — | — | — | — | |
| Flying Bird+Aug.=L1, Pruning Ratio=50%2026.01 | 69.31 | — | — | — | — | — | — | 21 | — | — | |
| DUQBackbone=ResNet-18, Cutout=false2020.12 | 69.01 | 0.5059 | — | — | — | — | — | — | — | — | |
| Deep EnsemblesBackbone=WRN-28-10, Cutout=true2020.12 | 68.75 | 0.1414 | — | — | — | — | — | — | — | — | |
| LG-FedAvgBackbone=ResNet-18, Initialization=Random2024.04 | 68.55 | — | — | — | — | — | — | — | — | — | |
| Flying Bird+Aug.=L1, Pruning Ratio=70%2026.01 | 68.4 | — | — | — | — | — | — | 10 | — | — | |
| TENTBackbone=ViT-B/162024.06 | 68 | — | — | — | — | — | — | — | — | — | |
| TENT2024.06 | 67.56 | — | — | — | — | — | — | — | — | — | |
| SAR2025.09 | 66.8 | — | — | — | — | — | — | — | — | — | |
| SHOT2025.09 | 66.8 | — | — | — | — | — | — | — | — | — | |
| FedAvg + FTBackbone=ResNet-18, Initialization=Random, Fine-tuning=true2024.04 | 66.65 | — | — | — | — | — | — | — | — | — | |
| Tent2025.09 | 66.5 | — | — | — | — | — | — | — | — | — |