ResearchTasksText-to-image concept steeringText-to-image concept steeringFollowBenchmarksDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyDatasetSOTA methodMetricTrendResultsLast updated167 held-out concepts (test)Prompting83.1Stage 1 Input Fidelity (VLM)4May 12, 2026