ResearchDatasetsBASALTFollowBenchmarksTask NameDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyTask NameDataset NameSOTA ResultTrendResultsHeld-out GeneralizationBASALT Top 10% hardest (n=30 clips)0.903Latency (s)3Held-out GeneralizationBASALT Top 25% hardest (n=75 clips)0.81Latency3Held-out GeneralizationBASALT Mean (n=300 human clips (100/task))0.629Latency3