Linjie Li

Learn how new methods improve models' visual and textual connections.

A new benchmark aims to assess MLLMs in video understanding across multiple topics.

New approach generates high-quality human action videos with depth information.