ResearchDatasetsHumanEvalFixFollowBenchmarksTask NameDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyTask NameDataset NameSOTA ResultTrendResultsCode RepairHumanEvalFix (test)59.1Success Rate (Python)19Bug-fixingHumanEvalFix held-out 25% (eval)5.2Utility Difference (Δπ)3