Image Classification on ImageNet-C (all 15 corruptions, severity 5)
61.36Top-1 AccuracyGrad
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| GradBackbone=ViTBase-LN2026.06 | 61.36 | — | 81.45 | |
| SVGDBackbone=ViTBase-LN2026.06 | 60.84 | — | 80.3 | |
| NaiveBackbone=ViTBase-LN2026.06 | 60.56 | — | 79.23 | |
| KLBackbone=ViTBase-LN2026.06 | 60.37 | — | 80.42 | |
| SARBackbone=ViTBase-LN2026.06 | 57.78 | — | 78.14 | |
| DeYOBackbone=ViTBase-LN2026.06 | 57.05 | — | 78.35 | |
| SARBackbone=ResNet-50-GN2026.06 | 38.12 | — | 59.7 | |
| GradBackbone=ResNet-50-GN2026.06 | 34.36 | — | 56.21 | |
| NaiveBackbone=ResNet-50-GN2026.06 | 33.06 | — | 55.91 | |
| SVGDBackbone=ResNet-50-GN2026.06 | 32.71 | — | 55.79 | |
| KLBackbone=ResNet-50-GN2026.06 | 32.66 | — | 52.23 | |
| TentBackbone=ViTBase-LN2026.06 | 32.36 | — | 47.4 | |
| DeYOBackbone=ResNet-50-GN2026.06 | 31.36 | — | 50.78 | |
| No AdaptBackbone=ResNet-50-GN2026.06 | 30.61 | — | 53.89 | |
| No AdaptBackbone=ViTBase-LN2026.06 | 29.94 | — | 54.21 | |
| TentBackbone=ResNet-50-GN2026.06 | 29.8 | — | 30.65 | |
| BETAEvaluation API=Clarifai, Corruption Severity=5, Budget per domain=$52026.04 | — | 54.7 | — | |
| LAMEEvaluation API=Clarifai, Corruption Severity=5, Budget per domain=$52026.04 | — | 43.2 | — | |
| SourceEvaluation API=Clarifai, Corruption Severity=5, Budget per domain=$52026.04 | — | 44.3 | — |