LLM-Driven On-Demand AI Service Provision in 6G Networks

Nan Cheng, Yifan Guo, Tingting Yang, Jinglong Shen, Wei Quan, Tom H. Luan

Abstract: With the continuous breakthroughs in the technical capabilities of large language models (LLMs), they have demonstrated exceptional technical efficacy across multiple application domains. Particularly in the field of task planning, this technological framework has proven capable of delivering systemic performance enhancements to communication systems. This paper focuses on the integrated application of LLMs in 6G network systems, adopting a hierarchical and progressive research paradigm for discussion. First, the technical advantages of LLMs are systematically elaborated, and a comparative analysis method is employed to deconstruct the functional demand disparities between the user side and the network side in 6G communication networks. Subsequently, an innovative LLM-driven on-demand artificial intelligence service architecture is proposed. Building upon this framework, simulation modeling is utilized to quantitatively evaluate its optimization effects on key performance metrics of communication networks. Finally, in the case study, the service architecture is empirically validated for its capability to meet differentiated network and user demands.

External IDs:doi:10.1109/mnet.2025.3627511