ResearchBenchmarksStructured video reasoning on Maze HardStructured video reasoning on Maze HardFollow81Exact Match RateTBS-2015.79232.72149.6566.579Jun 11, 2026Evaluation resultsMethodMethodLinksExact Match RateTBS-20Budget=20Budget=202026.0681TBS-10Budget=10Budget=102026.0668.7BoN-20Budget=20Budget=202026.0660BoN-10Budget=10Budget=102026.0653BoN-5Budget=5Budget=52026.0640BoN-1Budget=1Budget=12026.0618.3