ResearchBenchmarksImage Classification on ImageNet (test) (Clean and Attack Accuracy Metrics)For AI assistantsMarkdownContent indexImage Classification on ImageNet (test) (Clean and Attack Accuracy Metrics)Follow76.1Clean AccuracyNormal52.90858.92964.9570.971Feb 9, 2022Evaluation resultsMethodMethodLinksClean AccuracyCAA3a AccuracyCAA3b AccuracyCAA3c AccuracySemantic Attacks (Rand.) AccuracySemantic Attacks (Sched.) AccuracyFull Attacks (Rand.) AccuracyFull Attacks (Sched.) AccuracyNormalTraining Setting=†Training Setting=†2022.0276.100031.220.600MadryTraining Setting=†, Co...Training Setting=†, Constraint=l_inf2022.0262.413.99.216.21497.12.8GAT-fTraining Setting=†Training Setting=†2022.026019.218.918.443.538.918.511.8Fast-ATTraining Setting=†, Co...Training Setting=†, Constraint=l_inf2022.0253.89.55.511.46.33.63.11