Dynamic Token Compression for Efficient Video Understanding through Reinforcement Learning

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Wang, Shida, Hua, YongXiang, Tao, Zhou, Cao, Haoyu, Xu, Linli
Format:	Preprint
Published:	2026
Subjects:	Computer Vision and Pattern Recognition
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Internet

https://arxiv.org/abs/2603.26365