ResearchDatasetsconfigurationsFollowBenchmarksTask NameDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyTask NameDataset NameSOTA ResultTrendResultsPre-training extrapolationconfigurations 100B loss-equivalent Table 20.2Mean Relative Error4Pre-training extrapolationconfigurations 300B single-phase Table 20.1Mean Relative Error4