Image Classification on Caltech101 (Adversarial Evaluation)
81.47Top-1 Accuracy (Clean)WISE-FT
Evaluation results
| Method | Links | ||
|---|---|---|---|
| WISE-FTBackbone=ViT-B/32, Pre-training Dataset=ImageNet, Evaluation Protocol=Zero-shot2026.03 | 81.47 | 0 | |
| FLYPBackbone=ViT-B/32, Pre-training Dataset=ImageNet, Evaluation Protocol=Zero-shot2026.03 | 80.77 | 0 | |
| TPGMBackbone=ViT-B/32, Pre-training Dataset=ImageNet, Evaluation Protocol=Zero-shot2026.03 | 80.49 | 0 | |
| CLIPBackbone=ViT-B/32, Evaluation Protocol=Zero-shot2026.03 | 80.47 | 0 | |
| Vanilla FTBackbone=ViT-B/32, Pre-training Dataset=ImageNet, Evaluation Protocol=Zero-shot2026.03 | 79.85 | 0 | |
| SPDBackbone=ViT-B/32, Pre-training Dataset=ImageNet, Evaluation Protocol=Zero-shot2026.03 | 79.8 | 0 | |
| GRACEBackbone=ViT-B/32, Pre-training Dataset=ImageNet, Evaluation Protocol=Zero-shot2026.03 | 79.23 | 49.3 | |
| PMG-AFTBackbone=ViT-B/32, Pre-training Dataset=ImageNet, Evaluation Protocol=Zero-shot2026.03 | 77.1 | 53.2 | |
| FAREBackbone=ViT-B/32, Pre-training Dataset=ImageNet, Evaluation Protocol=Zero-shot2026.03 | 76.97 | 45.8 | |
| LAATBackbone=ViT-B/32, Pre-training Dataset=ImageNet, Evaluation Protocol=Zero-shot2026.03 | 75.1 | 47.2 | |
| TeCoABackbone=ViT-B/32, Pre-training Dataset=ImageNet, Evaluation Protocol=Zero-shot2026.03 | 74.03 | 51.4 |