ResearchBenchmarksMulti-label Toxic Content Classification on Jigsaw-ML (Adversarial Robustness)Multi-label Toxic Content Classification on Jigsaw-ML (Adversarial Robustness)Follow71.7Attack Success RateAT1-unk70.61877.921585.22592.5285Apr 9, 2024Evaluation resultsMethodMethodLinksAttack Success RateAvg QueriesAvg Perturbed WordsAT1-unkTraining=AT1-unk, Sear...Training=AT1-unk, Search=unk, Time (s)=31062024.0471.791.0811.84No ATTraining=No AT, Search...Training=No AT, Search=unk, Time (s)=11412024.0498.7549.386.96