Vision-Centric Activation and Coordination for Multimodal Large Language Models

Yunnan Wang, Fan Lu, Kecheng Zheng +4

Multimodal large language models (MLLMs) integrate image features from visual encoders with LLMs, demonstrating advanced comprehension capabilities. However, mainstream MLLMs are s…