Image Classification on ImageNet V2 (Acc, Rob)
64.7AccuracyDBD
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DBDAttack=PGD-100, Epsilon=4/255, Backbone=CLIP-ViT-B/162026.06 | 64.7 | 97.3 | |
| R-TPTAttack=PGD-100, Epsilon=4/255, Backbone=CLIP-ViT-B/162026.06 | 63.9 | 40.2 | |
| CLIPAttack=PGD-100, Epsilon=4/255, Backbone=CLIP-ViT-B/162026.06 | 63.4 | 0 | |
| DBDAttack=PGD-100, Epsilon=4/255, Backbone=CLIP-ViT-B/322026.06 | 59.1 | 94.3 | |
| MTACategory=Tuning-free, Backbone=CLIP-ViT-B/32, Attack scenario=PGD-100 (eps=4.0)2026.06 | 58.1 | 18.7 | |
| R-TPTAttack=PGD-100, Epsilon=4/255, Backbone=CLIP-ViT-B/322026.06 | 58 | 34.3 | |
| R-TPTCategory=Tuning-based, Backbone=CLIP-ViT-B/32, Attack scenario=PGD-100 (eps=4.0)2026.06 | 57.8 | 31 | |
| CLIPAttack=PGD-100, Epsilon=4/255, Backbone=CLIP-ViT-B/322026.06 | 57.2 | 0 | |
| DBDAttack=PGD-10, Epsilon=1/255, Backbone=CLIP-ResNet502026.06 | 55.7 | 93 | |
| CLIPCategory=Baseline, Backbone=CLIP-ViT-B/32, Attack scenario=PGD-100 (eps=4.0)2026.06 | 55 | 0 | |
| MACCategory=Tuning-free, Backbone=CLIP-ViT-B/32, Attack scenario=PGD-100 (eps=4.0)2026.06 | 54.8 | 40.4 | |
| CLIPAttack=PGD-10, Epsilon=1/255, Backbone=CLIP-ResNet502026.06 | 54.7 | 0 | |
| R-TPTAttack=PGD-10, Epsilon=1/255, Backbone=CLIP-ResNet502026.06 | 54.7 | 41.6 | |
| TPT2026.06 | 54.1 | 0.2 | |
| SS-TPT2026.06 | 53.7 | 41.7 | |
| Ensemble2026.06 | 51.8 | 30 | |
| CLIP2026.06 | 51.5 | 0.1 | |
| R-TPT2026.06 | 51.5 | 37.5 | |
| TTCCategory=Tuning-free, Backbone=CLIP-ViT-B/32, Attack scenario=PGD-100 (eps=4.0)2026.06 | 49.2 | 4.1 | |
| TTCAttack=PGD-100, Epsilon=4/255, Backbone=CLIP-ViT-B/162026.06 | 48.9 | 16.1 | |
| DOC2026.06 | 48.5 | 6.1 | |
| TTC2026.06 | 47.5 | 6.3 | |
| TTCAttack=PGD-100, Epsilon=4/255, Backbone=CLIP-ViT-B/322026.06 | 41.3 | 21.4 | |
| TeCoAAttack=PGD-10, Epsilon=1/255, Backbone=CLIP-ResNet502026.06 | 39.6 | 21.7 | |
| TeCoAAttack=PGD-100, Epsilon=4/255, Backbone=CLIP-ViT-B/322026.06 | 32.4 | 7.3 |