Semantic Consistency of Vision Tokens: A Vision-Centric Perspective on Multimodal Large Language Models

arXiv:2611.15448 · cs.CV · Submitted 2026-11-11 · Read on arXiv

cs.CV

Submitted: 2026-11-11

Updated: 2026-11-11

Code: https://github.com/djukicn/sgt

Terminology

Sources

Related papers