多维度多智能体分组讨论的虚假新闻检测方法

doi:10.3778/j.issn.1002-8331.2408-0259

摘要/Abstract

摘要： 大语言模型技术的快速发展使其在虚假新闻检测任务中展现出巨大应用潜力。但由于虚假新闻检测的推理复杂度较高，现有方法难以准确实现推理判断。模拟人对新闻的多维度感知过程，提出了一种多维度多智能体分组讨论的虚假新闻检测框架。通过子任务分解将虚假新闻检测分解为三个单选形式的阅读理解子任务，降低了推理过程难度，并设计多智能体分组讨论策略执行子任务，最终通过软分类打分，总结对比不同观点的子任务结果完成检测分类。该方法在FAKENEWSAMT和CELEBRITY数据集中的F1值分别为0.943和0.878，相较于以往基于大模型的检测方法具有更优的检测性能。

关键词: 多智能体, 虚假新闻检测, 大语言模型, 自然语言处理

Abstract: The rapid development of large language model technology has shown great potential for application in fake news detection tasks. However, due to the high complexity of reasoning involved in fake news detection, existing methods struggle to accurately perform inferential judgments. This paper simulates the multi-dimensional perception of news by humans and proposes a fake news detection method based on multi-dimensional and multi-agent group discussions. This framework breaks down fake news detection into three multiple-choice reading comprehension sub-tasks, thereby reducing the difficulty of the reasoning process. It designs a multi-agent group discussion strategy to execute subtasks, and uses soft classification scoring to summarize and compare the subtask results of different viewpoints to complete the final classification. This method achieves F1 scores of 0.943 and 0.878 on the FAKENEWSAMT and CELEBRITY datasets, respectively, demonstrating superior detection performance compared to previous detection methods based on large language models.

Key words: multi-agent, fake news detection, large language model, natural language processing

许旻辰, 屈丹, 司念文, 彭思思, 刘云鹏. 多维度多智能体分组讨论的虚假新闻检测方法[J]. 计算机工程与应用, 2025, 61(22): 183-195.

XU Minchen, QU Dan, SI Nianwen, PENG Sisi, LIU Yunpeng. Multi-Dimensional and Multi-Agent Group Discussion Method for Fake News Detection[J]. Computer Engineering and Applications, 2025, 61(22): 183-195.

参考文献

[1] KALIYAR R K, GOSWAMI A, NARANG P. FakeBERT: fake news detection in social media with a BERT-based deep learning approach[J]. Multimedia Tools and Applications, 2021, 80(8): 11765-11788.
[2] MIN E X, RONG Y, BIAN Y T, et al. Divide-and-conquer: post-user interaction network for fake news detection on social media[C]//Proceedings of the ACM Web Conference 2022. New York: ACM, 2022: 1148-1158.
[3] SUN L, RAO Y, LAN Y Q, et al. HG-SL: jointly learning of global and local user spreading behavior for fake news early detection[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2023, 37(4): 5248-5256.
[4] 强子珊, 顾益军. 融合动态传播和社区结构的社交媒体谣言检测模型[J]. 计算机工程与应用, 2024, 60(18): 198-207.
QIANG Z S, GU Y J. Rumor detection model based on dynamic propagation and community structure[J]. Computer Engineering and Applications, 2024, 60(18): 198-207.
[5] HU B Z, SHENG Q, CAO J, et al. Bad actor, good advisor: exploring the role of large language models in fake news dete-ction[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2024, 38(20): 22105-22113.
[6] WEI J, WANG X Z, SCHUURMANS D, et al. Chain-of-thought prompting elicits reasoning in large language models[C]//Proceedings of the 36th International Conference on Neural Information Processing Systems, 2024: 24824-24837.
[7] ZHANG X, GAO W. Towards LLM-based fact verification on news claims with a hierarchical step-by-step prompting method[C]//Proceedings of the 13th International Joint Conference on Natural Language Processing and the 3rd Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics. Stroudsburg: ACL, 2023: 996-1011.
[8] HUANG Y, SUN L. Harnessing the power of ChatGPT in fake news: an in-depth exploration in generation, detection and explanation[J]. arXiv:2310.05046, 2023.
[9] 胡杨, 王啸. 什么是“真实”: 数字媒体时代受众对假新闻的认知与辨识[J]. 新闻记者, 2019(8): 4-14.
HU Y, WANG X. What is “truth”: audience’s cognition and identification of fake news in the digital media era[J]. Shanghai Journalism Review, 2019(8): 4-14.
[10] PéREZ-ROSAS V, KLEINBERG B, LEFEVRE A, et al. Automatic detection of fake news[C]//Proceedings of the 27th International Conference on Computational Linguistics, 2018: 3391-3401.
[11] PELRINE K, IMOUZA A, THIBAULT C, et al. Towards reliable misinformation mitigation: generalization, uncertainty, and GPT-4[C]//Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing. Stroudsburg: ACL, 2023: 6399-6429.
[12] CHEN C Y, SHU K. Can LLM-generated misinformation be detected?[J]. arXiv:2309.13788, 2023.
[13] JIANG B H, TAN Z, NIRMAL A, et al. Disinformation det-ection: an evolving challenge in the age of LLMs[J]. arXiv:2309.15847, 2023.
[14] LIU H, WANG W Y, LI H R, et al. TELLER: a trustworthy framework for explainable, generalizable and controllable fake news detection[J]. arXiv:2402.07776, 2024.
[15] LEITE J A, RAZUVAYEVSKAYA O, BONTCHEVA K, et al. Detecting misinformation with LLM-predicted credibility signals and weak supervision[J]. arXiv:2309.07601, 2023.
[16] WANG H R, SHU K. Explainable claim verification via knowledge-grounded reasoning with large language models[C]//Findings of the Association for Computational Linguistics: EMNLP 2023. Stroudsburg: ACL, 2023: 6288-6304.
[17] LI X Y, ZHANG Y F, MALTHOUSE E C. Large language model agent for fake news detection[J]. arXiv:2405.01593, 2024.
[18] GUO T C, CHEN X Y, WANG Y Q, et al. Large language model based multi-agents: a survey of progress and challenges[J]. arXiv:2402.01680, 2024.
[19] CHAN C M, CHEN W Z, SU Y S, et al. ChatEval: towards better LLM-based evaluators through multi-agent debate[J]. arXiv:2308.07201, 2023.
[20] DU Y L, LI S, TORRALBA A, et al. Improving factuality and reasoning in language models through multiagent debate[J]. arXiv:2305.14325, 2023.
[21] WANG Q N, WANG Z H, SU Y, et al. Rethinking the bounds of LLM reasoning: are multi-agent discussions the key? [J]. arXiv:2402.18272, 2024.
[22] LIANG T, HE Z W, JIAO W X, et al. Encouraging divergent thinking in large language models through multi-agent debate[J]. arXiv:2305.19118, 2023.
[23] WANG Y X, WU Z Y, YAO J F, et al. TDAG: a multi-agent framework based on dynamic task decomposition and agent generation[J]. arXiv:2402.10178, 2024.
[24] QIAN C, LIU W, LIU H Z, et al. ChatDev: communicative agents for software development[J]. arXiv:2307.07924, 2023.
[25] HONG S R, ZHUGE M C, CHEN J Q, et al. MetaGPT: meta programming for a multi-agent collaborative framework[J]. arXiv:2308.00352, 2023.
[26] XU Z L, YU C, FANG F, et al. Language agents with reinforcement learning for strategic play in the werewolf game[J]. arXiv:2310.18940, 2023.
[27] WANG S Z, LIU C, ZHENG Z L, et al. Avalon’s game of thoughts: battle against deception through recursive contemplation[J]. arXiv:2310.01320, 2023.
[28] CHU Z, CHEN J C, CHEN Q L, et al. A survey of chain of thought reasoning: advances, frontiers and future[J]. arXiv:2309.15402, 2023.
[29] XIONG K, DING X, CAO Y X, et al. Examining inter-consistency of large language models collaboration: an in-depth analysis via debate[C]//Findings of the Association for Computational Linguistics: EMNLP 2023. Stroudsburg: ACL, 2023: 7572-7590.
[30] CHOUDHRY A, KHATRI I, JAIN M, et al. An emotion-aware multitask approach to fake news and rumor detection using transfer learning[J]. IEEE Transactions on Computational Social Systems, 2024, 11(1): 588-599.
[31] NAN Q, SHENG Q, CAO J, et al. Let silence speak: enhan-cing fake news detection with generated comments from large language models[C]//Proceedings of the 33rd ACM International Conference on Information and Knowledge Management. New York: ACM, 2024: 1732-1742.
[32] WAN H R, FENG S B, TAN Z X, et al. DELL: generating reactions and explanations for LLM-based misinformation detection[C]//Findings of the Association for Computational Linguistics: ACL 2024. Stroudsburg: ACL, 2024: 2637-2667.
[33] Anthropic. The Claude 3 model family: Opus, Sonnet, Haiku[EB/OL]. [2024-08-04]. https://www-cdn.anthropic.com/de8ba-9b01c9ab7cbabf5c33b80b7bbc618857627/Model_Card_Claude_
3.pdf.