EarlyTom: Early Token Compression Completes Fast Video Understanding

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Wang, Hesong, Jin, Xin, Lu, Lu, Li, Chenhaowen, Chen, Jian, Liu, Qiang, Wang, Huan
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/2605.30010