ResearchTasksSafety-aware fine-tuningSafety-aware fine-tuningFollowBenchmarksDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyDatasetSOTA methodMetricTrendResultsLast updatedREDORCA selected subset 2Online dynamic1.37Evaluation Loss9Jun 24, 2026DAHOAS RM-HH-RLHF (val)Online dynamic0.43Evaluation Loss9Jun 24, 2026REDORCA selected (subset 1)Online dynamic1.02Evaluation Loss9Jun 24, 2026BLUEORCA (val)Online dynamic0.82Evaluation Loss9Jun 24, 2026