V-SEAM: Visual Semantic Editing and Attention Modulating for Causal Interpretability of Vision-Language Models

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Wang, Qidong, Hu, Junjie, Jiang, Ming
Format:	Preprint
Published:	2025
Subjects:	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/2509.14837