大语言模型赋能智能网络的应用与挑战

摘要：大语言模型（LLM）正逐步融入网络的智能规划建设、智能维护、智能优化与网络运营等关键环节，在提升自动化与智能化水平方面展现出显著潜力。本文基于大模型与智能网络融合的背景，梳理了大模型在智能网络各关键领域的应用路径，总结其在提升决策效率、增强服务适配性、降低运维成本等方面的优势。深入探讨了智能网络环境下大模型面临的解空间组合爆炸与NP难（NP-hard）问题、多维度不确定性、实时性约束、数据异构性、人机协同与成本效益平衡等技术挑战，并归纳了现有应对思路。未来，随着多模态融合、在线学习与人机协同等技术的持续进步，大模型有望在推动网络从规则驱动向知识驱动转型的过程中发挥重要作用，为智能网络的发展提供新思路。

关键词：大语言模型；智能网络；人工智能；应用与挑战

Abstract: Large language models (LLMs) are gradually being integrated into key stages of intelligent network development, including network planning and construction, intelligent maintenance, optimization, and operations, demonstrating significant potential in enhancing automation and intelligence. This paper, grounded in the context of the convergence between LLMs and intelligent networks, reviews the application pathways of LLMs across critical areas of intelligent networks. It summarizes their advantages in improving decision-making efficiency, enhancing service adaptability, and reducing operational and maintenance costs. Furthermore, it explores the major technical challenges faced by LLMs in intelligent network environments, including the combinatorial explosion of solution spaces and NP-hard problems, multidimensional uncertainties, real-time constraints, data heterogeneity, human-machine collaboration, and cost-benefit trade-offs, and outlines current strategies for addressing these issues. Looking ahead, with the continued advancement of technologies such as multimodal integration, online learning, and human-machine collaboration, LLMs are expected to play an increasingly important role in facilitating the transition of networks from rule-driven to knowledge-driven paradigms, offering new perspectives for the development of intelligent networks.

Keywords: large language model; intelligent network; artificial intelligence; application and challenge