ResearchBenchmarksLanguage-conditioned visual reasoning on VLABenchFollow54SR (Toy)π022.830.93947.1Nov 26, 2025Evaluation ResultsMethodMethodLinksSR (Toy)SR (Fruit)SR (Painting)SR (Poker)SR (Mahjong)Average SRπ02025.1154481666.9826.2E02025.115434127218.7538.15π0 FAST2025.114642263020.8332.97π0.52025.11241836206.5220.9