大模型驱动的事实核查多智能体系统构建研究

打开文本图片集

中图分类号：TP18 文献标识码：A DOI：10.11968/tsyqb.1003-6938.2025046

AbstractAgainst the backdrop of globalization marked bypolitical and economic turbulence，intensifying climate change，andaccelerated technologicalinnovation，the informationdisseminationecosystem has grown increasinglycomplex.The proliferationofdisinformationhasemergedasanurgentisue threateningsocialtability，publichealth，andpolitical security.Firstofall，This paper focuses onthe innovative applicationof large languagemodel-drivenagents in the fieldof fact-checking toaddressthechalenges of disinformation governance incomplex scenarios.Inthenextplace，it systematically examines thecurrent challenges faced in fact-checkingand demonstrates the potential and advantages of empowering fact-checking work through large language model-drivenmulti-agent.This research designed the theoreticalframeworkofthelarge language model-drivenfact-checking multi-agentsystem，outlinedapractical path forcollaborative work among large language model-driven fact-checking multi-agent through empirical research，and verified the superior performance of the system through comparative experiments.

Key words disinformation; fact-checking; multi-agent; large language model

政治经济的不稳定、收入不平等的加剧、气候变化的严峻挑战、全球移民趋势的涌动，以及科技的迅猛发展，特别是人工智能技术的广泛应用，共同塑造了一个复杂多变的信息传播环境，导致虚假信息（Disinformation）和错误信息（Misinformation）爆炸式增长，在政治领域[1、健康议题[2以及社会秩序[3]方面造成了威胁，加剧了社会的不确定性和分裂趋势。（剩余18893字）