ResearchBenchmarksAdversarial Robustness on ImageNet-LTFor AI assistantsMarkdownContent indexAdversarial Robustness on ImageNet-LTFollow30.02Clean AccuracyFGSM-Our6.287212.448618.6124.7714Apr 6, 2021Evaluation resultsMethodMethodLinksClean AccuracyFGSM AccuracyPGD-20 AccuracyFGSM-Ourε=2/255ε=2/2552021.0430.0218.517.67FGSM-Ourε=4/255ε=4/2552021.0425.8813.4911.87FGSM-ATε=2/255ε=2/2552021.0425.6415.3214.59FGSM-ATε=4/255ε=4/2552021.0421.9410.889.45FAST-Ourε=2/255ε=2/2552021.0415.4511.5110.31FAST-ATε=2/255ε=2/2552021.0411.368.237.16FAST-Ourε=4/255ε=4/2552021.0410.767.286.13FAST-ATε=4/255ε=4/2552021.047.24.523.76