ResearchTasksReward ScoringFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast UpdatedHHA benchmarkSFT+DPO66.97Harmlessness Score (Base)30Mar 24, 2026NB-curated (evaluation set)TOMPA33.64Mean Reward Score6Apr 6, 2026