Jul 2026
VisualRouter: Query-Grounded Visual Sampling for Long Video Understanding
This paper proposes VisualRouter, a training-free and plug-and-play framework for query-grounded visual sampling, which consistently improves multiple LVLMs over uniform sampling, and outperforming existing training-free visual sampling methods under the same setting.
Haiyue Zhang, Yi Bin, Xun Jiang et al.
· arXiv.org · 0 citations