# Rethinking Generative Semantic Communication for Multi-User Systems with Large Language Models

arXiv: https://arxiv.org/abs/2408.08765 · HTML: https://mlsanity.com/paper/2408.08765/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 16 Aug 2024.

## Built on by (most cited first, 6 of 6)

- [Large Language Models for Wireless Communications: From Adaptation to Autonomy](https://mlsanity.com/paper/2507.21524/) (2025)
- [Generative Semantic Communication: Architectures, Technologies, and Applications](https://mlsanity.com/paper/2412.08642/) (2024)
- [Task-Oriented Multimodal Token Transmission in Resource-Constrained Multiuser Networks](https://mlsanity.com/paper/2505.07841/) (2025)
- [Semantic Edge-Cloud Communication for Real-Time Urban Traffic Surveillance with ViT and LLMs over Mobile Networks](https://mlsanity.com/paper/2509.21259/) (2025)
- [From Large AI Models to Agentic AI: A Tutorial on Future Intelligent Communications](https://mlsanity.com/paper/2505.22311/) (2025)
- [Large-Scale Model Enabled Semantic Communication Based on Robust Knowledge Distillation](https://mlsanity.com/paper/2508.02148/) (2025)

## Abstract (arXiv)

The surge in connected devices in 6G with typical complex tasks requiring multi-user cooperation, such as smart agriculture and smart cities, poses significant challenges to unsustainable traditional communication. Fortunately, the booming artificial intelligence technology and the growing computational power of devices offer a promising 6G enabler: semantic communication (SemCom). However, existing deep learning-based SemCom paradigms struggle to extend to multi-user scenarios due to its increasing model size with the growing number of users and its limited compatibility with complex communication environments. Consequently, to truly empower 6G networks with this critical technology, this article rethinks generative SemCom for multi-user system and proposes a novel framework called ``M-GSC" with the large language model (LLM) as the shared knowledge base (SKB). The LLM-based SKB plays three critical roles, that is, complex task decomposition, semantic representation specification, and semantic translation and mapping, for complex tasks, spawning a series of benefits such as semantic encoding standardization and semantic decoding personalization. Meanwhile, to enhance the performance of M-GSC framework, we highlight three optimization strategies unique to this framework: extending the LLM-based SKB into a multi-agent LLM system, offloading semantic encoding and decoding, and managing communication and computational resources. Finally, a case study is conducted to demonstrate the preliminary validation on the effectiveness of the M-GSC framework in terms of efficient decoding offloading.