ResearchDatasetsMT-Bench, HumanEval, and GSM8KFollowBenchmarksTask NameDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyTask NameDataset NameSOTA ResultTrendResultsSpeculative DecodingMT-Bench, HumanEval, and GSM8K Mean4.83Mean Acceptance Length (tau)26