List your company
ResearchFormerly WizWand

Image Classification on Oxfordpets (Robust Accuracy)

68.05Robust AccuracyOurs (5 trees)
-2.72215.651534.02552.3985Mar 19, 2026Mar 24, 2026Mar 29, 2026Apr 4, 2026Apr 9, 2026Apr 14, 2026Apr 20, 2026
Updated 6mo ago

Evaluation results

MethodLinks
Ours (5 trees)Zero-shot=true, Adversarial fine-tuning=ImageNet, PGD attack steps=20, Perturbation radius (epsilon)=1/255, Number of trees=52026.0468.05
OursZero-shot=true, Adversarial fine-tuning=ImageNet, PGD attack steps=20, Perturbation radius (epsilon)=1/2552026.0467.96
AoSZero-shot=true, Adversarial fine-tuning=ImageNet, PGD attack steps=20, Perturbation radius (epsilon)=1/2552026.0467.8
PMG-FTZero-shot=true, Adversarial fine-tuning=ImageNet, PGD attack steps=20, Perturbation radius (epsilon)=1/2552026.0463.07
TeCoAZero-shot=true, Adversarial fine-tuning=ImageNet, PGD attack steps=20, Perturbation radius (epsilon)=1/2552026.0462.8
Comp-TGAProtocol=Zero-shot, Attack=PGD, Epsilon=1/255, Training Dataset=Tiny-ImageNet2026.0360.09
TGA-ZSRProtocol=Zero-shot, Attack=PGD, Epsilon=1/255, Training Dataset=Tiny-ImageNet2026.0357.75
FAREZero-shot=true, Adversarial fine-tuning=ImageNet, PGD attack steps=20, Perturbation radius (epsilon)=1/2552026.0456.7
Comp-TGAMode=Zero-shot, Attack=Adaptive Attack, Epsilon=1/255, Training Dataset=Tiny-ImageNet2026.0356.06
TGA-ZSRMode=Zero-shot, Attack=Adaptive Attack, Epsilon=1/255, Training Dataset=Tiny-ImageNet2026.0354.62
Comp-TGAZero-shot=true, Attack=APGD (epsilon=1/255)2026.0341.91
PMG-AFTProtocol=Zero-shot, Attack=PGD, Epsilon=1/255, Training Dataset=Tiny-ImageNet2026.0341.62
TGA-ZSRZero-shot=true, Attack=APGD (epsilon=1/255)2026.0341.13
PMG-AFTZero-shot=true, Attack=APGD (epsilon=1/255)2026.0340.31
PMG-AFTMode=Zero-shot, Attack=Adaptive Attack, Epsilon=1/255, Training Dataset=Tiny-ImageNet2026.0339.85
TeCoAZero-shot=true, Attack=APGD (epsilon=1/255)2026.0336.5
TeCoAMode=Zero-shot, Attack=Adaptive Attack, Epsilon=1/255, Training Dataset=Tiny-ImageNet2026.0336.44
TeCoAProtocol=Zero-shot, Attack=PGD, Epsilon=1/255, Training Dataset=Tiny-ImageNet2026.0336.44
LAATZero-shot=true, Attack=APGD (epsilon=1/255)2026.0336
LAATProtocol=Zero-shot, Attack=PGD, Epsilon=1/255, Training Dataset=Tiny-ImageNet2026.0335.54
LAATMode=Zero-shot, Attack=Adaptive Attack, Epsilon=1/255, Training Dataset=Tiny-ImageNet2026.0335
FT-Adv.Protocol=Zero-shot, Attack=PGD, Epsilon=1/255, Training Dataset=Tiny-ImageNet2026.0334.97
FT-Adv.Zero-shot=true, Attack=APGD (epsilon=1/255)2026.0334.86
FT-Adv.Mode=Zero-shot, Attack=Adaptive Attack, Epsilon=1/255, Training Dataset=Tiny-ImageNet2026.0334.07
Comp-TGAZero-shot=true2026.0328.41
FAREZero-shot=true, Attack=APGD (epsilon=1/255)2026.0327.8
FAREProtocol=Zero-shot, Attack=PGD, Epsilon=1/255, Training Dataset=Tiny-ImageNet2026.0324.56
TGA-ZSRZero-shot=true2026.0324.5
FAREMode=Zero-shot, Attack=Adaptive Attack, Epsilon=1/255, Training Dataset=Tiny-ImageNet2026.0324.45
PMG-AFTZero-shot=true2026.0320.68
LAATZero-shot=true2026.0319.22
TeCoAZero-shot=true2026.0316.61
FT-Adv.Zero-shot=true2026.0316.45
FAREZero-shot=true2026.039.67
CLIPProtocol=Zero-shot, Attack=PGD, Epsilon=1/255, Training Dataset=Tiny-ImageNet2026.033.84
CLIPZero-shot=true, Adversarial fine-tuning=ImageNet, PGD attack steps=20, Perturbation radius (epsilon)=1/2552026.043.79
CLIPZero-shot=true2026.032.97
FT-CleanProtocol=Zero-shot, Attack=PGD, Epsilon=1/255, Training Dataset=Tiny-ImageNet2026.032.4
FT-CleanZero-shot=true2026.032.03
FT-CleanZero-shot=true, Attack=APGD (epsilon=1/255)2026.030.03
CLIPZero-shot=true, Attack=APGD (epsilon=1/255)2026.030