Research on Cloud-Edge Computing for LLM

Large Language Models (LLMs) offer remarkable capabilities across various natural language processing (NLP) tasks. However, their high computational complexity and significant memory requirements during the training and inference stages hinder their practical application. Cloud-edge computing has emerged as a promising paradigm to overcome these limitations by enabling the collaborative execution of LLM inference across cloud and edge servers. This paper surveys recent research on cloud-edge collaborative frameworks for LLMs, with a focus on system architecture, optimization objectives, and learning-based offloading strategies.