What Semantics Survive the Connector? Diagnosing VLM-to-DiT Alignment in Video Editing

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Lin, Hangyu, Wen, Chao, Xu, Chengming, Gao, Jianxiong, Zhang, Jiangning, Hu, Xiaobin, Fu, Yanwei
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/2605.20795