Image Classification on Oxfordpets (Adversarial Robustness: Clean, PGD, CW, Auto Attack)
87.41Clean AccuracyCLIP
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| CLIP2025.12 | 87.41 | 0.64 | 0.87 | 0.03 | |
| FAREFine-tuning dataset=TinyImageNet, Adversarial training=2-step PGD, Epsilon=1/2552025.12 | 80.98 | 10.93 | 15.15 | 0.93 | |
| PMG-AFTFine-tuning dataset=TinyImageNet, Adversarial training=2-step PGD, Epsilon=1/2552025.12 | 79.4 | 44.59 | 28.05 | 18.94 | |
| UCATFine-tuning dataset=TinyImageNet, Adversarial training=2-step PGD, Epsilon=1/2552025.12 | 74.16 | 45.11 | 45.38 | 44.02 | |
| TGA-ZSRFine-tuning dataset=TinyImageNet, Adversarial training=2-step PGD, Epsilon=1/2552025.12 | 68.44 | 39.96 | 40.2 | 38.84 | |
| TeCoAFine-tuning dataset=TinyImageNet, Adversarial training=2-step PGD, Epsilon=1/2552025.12 | 61.3 | 37.8 | 37.37 | 36.58 |