ResearchTasksJudge AgreementFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast UpdatedMT-bench Second Turn 1.0GPT-4 (Pairwise)95Agreement Rate46Feb 26, 2026MT-bench Second TurnG4-Pair95Agreement32Feb 26, 2026Chatbot Arena Random = 50% (S2)GPT-3.5 (Pairwise)96Agreement10Feb 26, 2026Chatbot Arena Random = 33% (S1)GPT-4 (Pairwise)72Agreement Rate10Feb 26, 2026