Image Classification on Flowers102 (Adversarial Robustness)
65.52Accuracy (Clean)CLIP
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| CLIPFine-tuning dataset=TinyImageNet, Adversarial training=2-step PGD, Epsilon=1/255, Evaluation protocol=Zero-shot2025.12 | 65.52 | — | — | 1.11 | — | |
| CLIP2025.12 | 65.49 | 1.62 | 2.03 | 0.02 | — | |
| CLIPDefense Category=None2025.11 | 65.46 | — | — | — | 0 | |
| TTEDefense Category=Test-time Defense2025.11 | 65.2 | — | — | — | 3.52 | |
| ATACDefense Category=Test-time Defense2025.11 | 64.29 | — | — | — | 85.69 | |
| TTCDefense Category=Test-time Defense2025.11 | 63.24 | — | — | — | 13.6 | |
| R-TPTDefense Category=Test-time Defense2025.11 | 62.92 | — | — | — | 32.46 | |
| CLIP-FTDefense Category=Adversarial Finetuning2025.11 | 53.37 | — | — | — | 0.13 | |
| PMG-AFTFine-tuning dataset=TinyImageNet, Adversarial training=2-step PGD, Epsilon=1/2552025.12 | 48.43 | 26.43 | 17.53 | 12.59 | — | |
| FAREFine-tuning dataset=TinyImageNet, Adversarial training=2-step PGD, Epsilon=1/2552025.12 | 47.63 | 6.52 | 5.66 | 0.6 | — | |
| UCATFine-tuning dataset=TinyImageNet, Adversarial training=2-step PGD, Epsilon=1/255, Evaluation protocol=Zero-shot2025.12 | 47.57 | — | — | 24.54 | — | |
| UCATFine-tuning dataset=TinyImageNet, Adversarial training=2-step PGD, Epsilon=1/2552025.12 | 47.57 | 26.79 | 24.95 | 24.54 | — | |
| FAREDefense Category=Adversarial Finetuning2025.11 | 41.01 | — | — | — | 0.62 | |
| TGA-ZSRFine-tuning dataset=TinyImageNet, Adversarial training=2-step PGD, Epsilon=1/2552025.12 | 39.81 | 22.44 | 21.52 | 20.98 | — | |
| TeCoAFine-tuning dataset=TinyImageNet, Adversarial training=2-step PGD, Epsilon=1/2552025.12 | 32.04 | 19.17 | 18.12 | 17.81 | — | |
| PMG-AFTDefense Category=Adversarial Finetuning2025.11 | 28.88 | — | — | — | 4.26 | |
| TeCoADefense Category=Adversarial Finetuning2025.11 | 27.78 | — | — | — | 3.81 | |
| ACATFine-tuning dataset=TinyImageNet, Adversarial training=2-step PGD, Epsilon=1/255, Evaluation protocol=Zero-shot2025.12 | 26.8 | — | — | 16.56 | — | |
| DKLFine-tuning dataset=TinyImageNet, Adversarial training=2-step PGD, Epsilon=1/255, Evaluation protocol=Zero-shot2025.12 | 25.94 | — | — | 16.23 | — | |
| TRADESFine-tuning dataset=TinyImageNet, Adversarial training=2-step PGD, Epsilon=1/255, Evaluation protocol=Zero-shot2025.12 | 24.48 | — | — | 16.46 | — |