AnalysisAI ModelsNovember 17, 2026

VideoNexus framework improves long-video perception and reasoning

Read original source →arxiv.org

VideoNexus separates video perception from evidence-driven planning using sub-agents to return compact, temporally grounded evidence. The framework achieves higher accuracy at lower cost on benchmarks including LVBench and Video-GAIA.

1 source

More stories today

Open the live feed