Towards Intelligence-Native Communication: ChatGLM-Assisted Multimodal Semantic Coding Paradigm

Semantic communication (SC) has emerged as a crucial enabler for task-oriented intelligent communication in sixth-generation (6G) networks. Despite recent advancements, existing multimodal SC systems remain challenging due to semantic heterogeneity and limited generalization of deep learning-based models under dynamic task and channel conditions. To address these limitations, we propose a ChatGLM-assisted multimodal semantic encoding and decoding framework. Our design follows a three-tier collaborative paradigm consisting of modality alignment, semantic abstraction, and representation adaptation. Specifically, a semantic alignment engine built upon ImageBind is employed to mitigate cross-modal discrepancies, while a bidirectional projection mechanism is further designed to enable seamless semantic conversion between neural feature spaces and ChatGLM-interpretable representations. Extensive experiments on two representative tasks demonstrate the effectiveness of the proposed framework, highlighting its potential to advance the development of next-generation multimodal SC.