Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Large Language Models for Wireless Communications: From Adaptation to Autonomy

Le Liang, Hao Ye, Yucheng Sheng, Ouya Wang, Jiacheng Wang, Shi Jin, Geoffrey Ye Li

Abstract

The emergence of large language models (LLMs) has revolutionized artificial intelligence, offering unprecedented capabilities in reasoning, generalization, and zero-shot learning. These strengths open new frontiers in wireless communications, where increasing complexity and dynamics demand intelligent and adaptive solutions. This article explores the role of LLMs in transforming wireless systems across three key directions: adapting pretrained LLMs for communication tasks, developing wireless-specific foundation models to balance versatility and efficiency, and enabling agentic LLMs with autonomous reasoning and coordination capabilities. We highlight recent advances, practical case studies, and the unique benefits of LLM-based approaches over traditional methods. Finally, we outline open challenges and research opportunities, including multimodal fusion, collaboration with lightweight models, and self-improving capabilities, charting a path toward intelligent, adaptive, and autonomous wireless networks.

Large Language Models for Wireless Communications: From Adaptation to Autonomy

Abstract

The emergence of large language models (LLMs) has revolutionized artificial intelligence, offering unprecedented capabilities in reasoning, generalization, and zero-shot learning. These strengths open new frontiers in wireless communications, where increasing complexity and dynamics demand intelligent and adaptive solutions. This article explores the role of LLMs in transforming wireless systems across three key directions: adapting pretrained LLMs for communication tasks, developing wireless-specific foundation models to balance versatility and efficiency, and enabling agentic LLMs with autonomous reasoning and coordination capabilities. We highlight recent advances, practical case studies, and the unique benefits of LLM-based approaches over traditional methods. Finally, we outline open challenges and research opportunities, including multimodal fusion, collaboration with lightweight models, and self-improving capabilities, charting a path toward intelligent, adaptive, and autonomous wireless networks.

Paper Structure

This paper contains 20 sections, 6 figures.

Table of Contents

  1. Introduction
  2. Adapting LLMs to Wireless Tasks
  3. Physical Layer Prediction Tasks
  4. Resource Allocation
  5. Semantic Communication
  6. Case Study: BP-LLM for Robust Beam Prediction
  7. Wireless Foundation Model
  8. Physical Layer Foundation Models
  9. Predictive Foundation Models for Resource Management
  10. Case Study: Zero-Shot Generalization in Multi-Task Predictive Foundation Model
  11. Agentic LLM for Wireless Communication
  12. Key Capabilities of Agentic LLMs
  13. Applications in Wireless Systems
  14. Case Study: Multi-AP Coordination with Agentic LLMs
  15. Challenges and Opportunities
  16. ...and 5 more sections

Figures (6)

  • Figure 1: Network architecture of LLM-based beam prediction.
  • Figure 2: Illustration of personalized semantic extraction via LLM in the semantic communication system for group photo transmission. Currently, Justin (sender) transmits this photo to Jane (receiver), and both of them intend to identify people and background in this photo. In the sender side, the LLM uses their prompt bases to extract intention-aligned semantics (omitting irrelevant details like clothing colors). Furthermore, the LLM transforms sender-centric references to receiver-perspective semantics ("me" to "Justin", "Jane" to "me"). Due to their intention only on people and background, clothing colors in the recovered image differ from the original. New users David and Helen can seamlessly join in this system by providing their prompts focused on clothing. The LLM now prioritizes clothes details over background, yielding a recovered image for Helen that preserves original clothing colors but alters the background.
  • Figure 3: Beam prediction performance of the BP-LLM method Sheng2025beam compared with other learning-based baselines under the mismatched scenarios.
  • Figure 4: The zero-shot prediction performance of the wireless foundation model compared with the full-shot baselines under the user distance prediction task.
  • Figure 5: Agentic workflow and coordination protocol for multi-AP cooperation.
  • ...and 1 more figures