ResearchDatasetsHARMAMPFollowBenchmarksTask NameDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyTask NameDataset NameSOTA ResultTrendResultsLLM Safety EvaluationHARMAMP Benign instances8.9Over-Refusal Rate30LLM Safety EvaluationHARMAMP (malicious instances)86.36Harm Score30