ResearchTasksBigram Language ModelingBigram Language ModelingFollowBenchmarksDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyDatasetSOTA methodMetricTrendResultsLast updatedSynthetic WebText initialization (val)Mixture KLs0.001Avg JS30Feb 26, 2026Synthetic Random 50% initialization (val)Mixture KLs0Avg JS Divergence30Feb 26, 2026