ResearchTasksSubword-to-byte distribution approximationFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast Updatedreference distribution token-to-byte 2025a (test)pgpt2 ◦ fα0Average JSD / byte36Mar 6, 2026