Cihang Xie

New benchmark assesses how video-language models handle inaccuracies effectively.

A model that improves segmentation of parts and objects in images.

A framework using memory tokens improves video understanding and interaction.