Image Classification on ImageNet-100 (test) Robustness
91.14Top-1 Accuracy (Original)SS-CA
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| SS-CABackbone=CLIP (ViT /32b)2025.11 | 91.14 | 75.66 | 88.84 | 85.68 | 86.48 | 72.3 | 91 | |
| Chen et al.Backbone=CLIP (ViT /32b)2025.11 | 89.83 | 74.83 | 88.17 | 84.69 | 85.32 | 70.98 | 90.12 | |
| Xiao et al.Backbone=CLIP (ViT /32b)2025.11 | 89.77 | 73.38 | 88.32 | 84.2 | 85.06 | 70.27 | 89.95 | |
| Conventional TrainingBackbone=CLIP (ViT /32b)2025.11 | 89.5 | 72.76 | 87.9 | 83.86 | 84.8 | 69.6 | 89.62 |