ResearchDatasetsVisual UnderstandingFollowBenchmarksTask NameDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyTask NameDataset NameSOTA ResultTrendResultsVisual UnderstandingVisual Understanding (VQAv2, NLVR2, MME) (held-out)71.54Accuracy4Visual Question AnsweringVisual-Understanding82.4Accuracy2