ResearchTasksSkill predictionFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast UpdatedSkillsBench Real-task holdout n=65TF-IDF74.2Set F110Jul 1, 2026SkillsBench Synthetic n=494 (test)SkillComposer73.9Set F110Jul 1, 2026