In recent years, Large Language Models (LLMs) have demonstrated remarkable capabilities across diverse NLP tasks. Extensive research has explored how to enhance the logical reasoning abilities such as Chain-of-Thought, Chain-of-Thought with Self-Consistency, Tree-Of-Thoughts, and multi-agent debates. In the context of multi-agent debates, significant performance improvements can be achieved with an increasing number of agents and debate rounds. However, the escalation in the number of agents and debate rounds can drastically raise the tokens cost of debates, thereby limiting the scalability of the multi-agent debate technique. To better harness the advantages of multi-agent debates in logical reasoning tasks, this paper proposes a method to significantly reduce token cost in multi-agent debates. This approach involves dividing all agents into multiple debate groups, with agents engaging in debates within their respective groups and sharing interim debate results between groups. Comparative experiments across multiple datasets have demonstrated that this method can reduce the total tokens by up to 51.7% during debates and while potentially enhancing accuracy by as much as 25%. Our method significantly enhances the performance and efficiency of interactions in the multi-agent debate.
Gradual Vigilance and Interval Communication: Enhancing Value Alignment in Multi-Agent Debates
National Engineering Research Center of Educational Big Data, Central China Normal University（华中师范大学教育大数据国家工程研究中心）；Faculty of Artificial Intelligence in Education, Central China Normal University（华中师范大学人工智能教育学部）
Beyond Single Models: Enhancing LLM Detection of Ambiguity in Requests through Debate
Institutes of Innovation for Future Society（未来社会创新研究所）；Nagoya University（名古屋大学）；Department of Micro-Nano Mechanical Science and Engineering（微纳米机械科学与工程系）
Self-signals Driven Multi-LLM Debate for Efficient and Accurate Reasoning
University of Cambridge（剑桥大学）；Sorbonne Université（索邦大学）；University of Science and Technology of China（中国科学技术大学）；Beihang University（北航大学）；Nanyang Technological University（南洋理工大学）
MAD-Fact: A Multi-Agent Debate Framework for Long-Form Factuality Evaluation in LLMs
Institute of Information Engineering, Chinese Academy of Sciences, Beijing 100085, China（中国科学院信息工程研究所）；School of Cyber Security, University of Chinese Academy of Sciences, Beijing 100049, China（中国科学院大学网络安全学院）
Enhancing Diagnostic Accuracy through Multi-Agent Conversations: Using Large Language Models to Mitigate Cognitive Bias
Singapore General Hospital（新加坡中央医院）；Duke-NUS Medical School（杜克-新加坡国立大学医学院）；Singapore Eye Research Institute（新加坡眼科研究所）；Singapore National Eye Center（新加坡国家眼科中心）；Byers Eye Institute, Stanford University（斯坦福大学拜尔斯眼科研究所）；National University of Singapore（新加坡国立大学）
A Few Hypocrites: Few-Shot Learning and Subtype Definitions for Detecting Hypocrisy Accusations in Online Climate Change Debates
Hertie School（赫蒂学院）；Hebrew University of Jerusalem（耶路撒冷希伯来大学）；University of Hamburg（汉堡大学）；Ilmenau University of Technology（伊尔默瑙工业大学）；University of Zurich（苏黎世大学）；Vrije Universiteit Amsterdam（阿姆斯特丹自由大学）
Breaking the Martingale Curse: Multi-Agent Debate via Asymmetric Cognitive Potential Energy
Mohamed bin Zayed University of Artificial Intelligence（莫扎德·本·扎耶德人工智能大学）；Gaoling School of Artificial Intelligence, Renmin University of China（中国人民大学国际关系学院）；Harvard University（哈佛大学）
Evaluation and Facilitation of Online Discussions in the LLM Era: A Survey
Archimedes, Athena Research Center（阿希米德研究中心）；Athens University of Economics and Business（雅典经济与商业大学）；École Normale Supérieure Paris-Saclay（巴黎萨克雷高等师范学校）；INSEAD
DEBISS: a Corpus of Individual, Semi-structured and Spoken Debates
Federal University of Campina Grande (UFCG) - System（坎皮纳格雷联邦大学（UFCG）- 系统）；Computing Campina Grande, Brazil Federal Institute of Paraíba (IFPB) Monteiro, Brazil（坎皮纳格雷计算，巴西帕拉州联邦理工学院（IFPB）蒙特伊罗，巴西）
Group discussions are essential for organizing every aspect of modern life, from faculty meetings to senate debates, from grant review panels to papal conclaves. While costly in terms of time and organization effort, group discussions are commonly seen as a way of reaching better decisions compared to solutions that do not require coordination between the individuals (e.g. voting)---through discussion, the sum becomes greater than the parts. However, this assumption is not irrefutable: anecdotal evidence of wasteful discussions abounds, and in our own experiments we find that over 30% of discu…
Chain-of-Discussion: A Multi-Model Framework for Complex Evidence-Based Question Answering
Wangxuan Institute of Computer Technology, Peking University（北京大学王选计算机研究所）；Center for Data Science, Peking University（北京大学数据科学中心）；Key Laboratory of Intelligent Press Media Technology, Peking University（北京大学智能新闻媒体技术重点实验室）
St. John’s College（圣约翰学院）；Carnegie Mellon University（卡内基梅隆大学）；Santa Fe Institute（圣塔菲研究所）
We present a structured random-walk model that captures key aspects of how people communicate in groups. Our model takes the form of a correlated Lévy flight that quantifies the balance between focused discussion of an idea and long-distance leaps in semantic space. We apply our model to three cases of increasing structural complexity: philosophical texts by Aristotle, Hume, and Kant; four days of parliamentary debate during the French Revolution; and branching comment trees on the discussion website Reddit. In the philosophical and parliamentary cases, the model parameters that describe this …
Peking University（北京大学）；Alibaba Group（阿里巴巴集团）；Indiana University Bloomington（印第安纳大学伯明顿分校）
Legal judgment prediction(LJP) is an essential task for legal AI. While prior methods studied on this topic in a pseudo setting by employing the judge-summarized case narrative as the input to predict the judgment, neglecting critical case life-cycle information in real court setting could threaten the case logic representation quality and prediction correctness. In this paper, we introduce a novel challenging dataset from real courtrooms to predict the legal judgment in a reasonably encyclopedic manner by leveraging the genuine input of the case -- plaintiff's claims and court debate data, fr…
