Image Classification on ImageNet-C distribution shifted variants
64.6Average AccuracyEATA
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| EATAType=BP-based, Backbone=ViT-Base2026.03 | 64.6 | — | — | — | — | — | |
| DeYOType=BP-based, Backbone=ViT-Base2026.03 | 63.9 | — | — | — | — | — | |
| ZOTTAType=BP-free, Backbone=ViT-Base2026.03 | 62.8 | — | — | — | — | — | |
| FOAType=BP-free, Backbone=ViT-Base2026.03 | 60.8 | — | — | — | — | — | |
| SARType=BP-based, Backbone=ViT-Base2026.03 | 60.7 | — | — | — | — | — | |
| TENTType=BP-based, Backbone=ViT-Base2026.03 | 60.4 | — | — | — | — | — | |
| NoAdaptType=/, Backbone=ViT-Base2026.03 | 55.5 | — | — | — | — | — | |
| DEnsBackbone=ResNet-502022.07 | 50.05 | 66 | 56 | 47 | 36 | 22 | |
| Δ-UQBackbone=ResNet-502022.07 | 48.95 | 61.7 | 53.1 | 44.2 | 33.2 | 21.8 | |
| SVIBackbone=ResNet-502022.07 | 48.05 | 63 | 53 | 43 | 31 | 20 | |
| VanillaBackbone=ResNet-502022.07 | 47.8 | 62.5 | 52 | 42 | 30 | 19.5 | |
| MC DropoutBackbone=ResNet-502022.07 | 46 | 60 | 50 | 38 | 29 | 17 |