ResearchDatasetsHarmful HEx-PHIHarmful HEx-PHIFollowBenchmarksTaskDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyTaskDatasetSOTA resultTrendResultsFinetuning attack successHarmful HEx-PHI (adversarial)84.92ASR9Safety Divergence MeasurementHarmful HEx-PHI1.3Max |h|1