ResearchTasksLLM Chat EvaluationFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast UpdatedArenaHardOfficial Instruct Model49.2Accuracy17Jun 2, 2026Arena-Hard v0.1 (test)Ours40.1Win Rate6Mar 4, 2026