A Trustworthy Agentic Multi-LLM Network: Challenges, Solutions, and a Use Case

Large Language Models (LLMs) demonstrate strong potential across a variety of tasks in communications and networking due to their advanced reasoning capabilities. However, since different LLMs have different model structures and are trained using distinct corpora and methods, they may offer varying optimization strategies for the same network issues. Moreover, the potential maliciousness of its hosting device can result in LLM responses with low confidence or even bias. To address these challenges, we propose a collaborative framework that organizes distributed LLMs into an Agentic Multi-LLM Network (Agentic MultiLLMN). This novel architecture transforms individual network nodes into collaborative autonomous agents equipped with LLMs. Through agentic functionalities such as autonomous planning, reasoning, and collaborative decision-making, the framework can provide high-quality responses for complex network optimization problems. Specifically, we first review related work and highlight the limitations of existing LLMs in collaboration and trust. We then introduce a workflow of the proposed Trustworthy Agentic MultiLLMN framework, which leverages blockchain to provide a verifiable audit trail for autonomous agent interactions. Given the severity of False Base Station (FBS) attacks in wireless communications, we present FBS defense as a case study. In this scenario, each Legitimate Base Station (LBS) acts as an agentic LLM node, collaborating to optimize power allocation and mitigate attacks. The simulation shows that this framework can more effectively resist this attack compared to other solutions. Finally, we outline promising future research directions in this emerging area.