ResearchBenchmarksVisual Question Answering on VQAv2 Reliable Challenge (test-dev)For AI assistantsMarkdownContent indexVisual Question Answering on VQAv2 Reliable Challenge (test-dev)Follow83.99AccuracyDAVR79.790581.8902583.9986.08975Dec 16, 2025Evaluation resultsMethodMethodLinksAccuracyPhi100AUC (100x)Coverage @ 0.5%Coverage @ 1%Coverage @ 2%Coverage @ 5%DAVR2025.1283.9937.5297.1544.5653.0363.4977.97