ResearchDatasetsHarmful HEx-PHIFollowBenchmarksTask NameDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyTask NameDataset NameSOTA ResultTrendResultsFinetuning attack successHarmful HEx-PHI (adversarial)84.92ASR9Safety Divergence MeasurementHarmful HEx-PHI1.3Max |h|1