M3LLM: Model Context Protocol-aided Mixture of Vision Experts For Multimodal LLMs in Networks

Yongjie Zeng, Hongyang Du

Current Multimodal Large Language Models (MLLMs) rely on centralized architectures and often suffer from poor alignment between the input task and their fixed visual encoding modul…