ResearchTasksGeneral LLM accelerationFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast UpdatedLongBench, InfiniteBench, RULER, AIME, MATH500 AveragePRR1.56Average Decoding Speedup12Jun 30, 2026