VLMs Need Words: Vision Language Models Ignore Visual Detail In Favor of Semantic Anchors

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Shahgir, Haz Sameen, Chen, Xiaofu, Fu, Yu, Shayegani, Erfan, Abu-Ghazaleh, Nael, Kementchedjhieva, Yova, Dong, Yue
Format:	Preprint
Published:	2026
Subjects:	Computer Vision and Pattern Recognition Computation and Language
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Internet

https://arxiv.org/abs/2604.02486