MNC: A multi-agent framework for complex network configuration

Recent progress in large language models (LLMs) has led to substantial improvements in their ability to perform a wide range of natural language processing tasks, particularly in handling complex scenarios. These models exhibit strong generalization capabilities and reasoning skills, making them well-suited for tasks that require integrating external knowledge and logical reasoning. In this paper, we introduce Multi-agent based Network Configuration (MNC), a novel multi-agent framework designed to leverage LLMs for complex network configuration tasks. Our framework consists of three core components: (1) the Requirement Analysis Module, which interprets user queries and retrieves relevant external network configuration knowledge; (2) the Configuration Generation Module, which uses an iterative Chain-of-Thought (COT) approach to produce and refine multiple analysis pathways; and (3) the Configuration Refinement Module, which evaluates and improves the final network configuration through a reflection-driven mechanism. We evaluate MNC on a network configuration dataset, where our proposed MNC outperforms existing baseline methods. Furthermore, an ablation study demonstrates the individual contributions of each module to the framework’s overall effectiveness. This research underscores the potential of LLM-based systems to advance complex network configuration tasks.