ResearchTasksCross-distribution performanceFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast UpdatedSAIR official benchmark (hard3)hard3_overfitted81.3Accuracy13Apr 23, 2026SAIR (hard2)98_hard20099Accuracy13Apr 23, 2026