ResearchTasksVisual Commonsense Reasoning (Q→A)Visual Commonsense Reasoning (Q→A)FollowBenchmarksDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyDatasetSOTA methodMetricTrendResultsLast updatedVCR Shortcut Mitigated Evaluation SM (test)VILLA (MAD*) [ROBERTa/CLIP-V]76.32Q->A Accuracy (100%)13Feb 26, 2026VCR Standard Evaluation (test)VILLA (MAD*) [ROBERTa/CLIP-V]61.49Q->A Accuracy (3%, 1000 SP/C)13Feb 26, 2026VCR (val)MERLOT79.4Accuracy8Feb 26, 2026VCR sampled 5000 (val)MerlotReserve0.84Accuracy7Jun 19, 2026
VCR Shortcut Mitigated Evaluation SM (test)VILLA (MAD*) [ROBERTa/CLIP-V]76.32Q->A Accuracy (100%)13Feb 26, 2026
VCR Standard Evaluation (test)VILLA (MAD*) [ROBERTa/CLIP-V]61.49Q->A Accuracy (3%, 1000 SP/C)13Feb 26, 2026