ResearchDatasetsTwenty QuestionsFollowBenchmarksTask NameDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyTask NameDataset NameSOTA ResultTrendResultsTwenty QuestionsTwenty Questions held-out (test)35.65Mean@810Reward PerformanceTwenty Questions High-data5.83Test-Time Reward8Reward PerformanceTwenty Questions Low-data3.46Test-Time Reward8