Revealing and Enhancing Core Visual Regions: Harnessing Internal Attention Dynamics for Hallucination Mitigation in LVLMs

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Lyu, Guangtao, Liu, Qi, Xu, Chenghao, Yan, Jiexi, Yang, Muli, Li, Xueting, Fang, Fen, Deng, Cheng
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/2602.15556