Multi-Agent Debate~(MAD) has emerged as a promising paradigm for improving the performance of large language models through collaborative reasoning. Despite recent advances, the key factors driving MAD's effectiveness remain unclear. In this work, we disentangle MAD into two key components--Majority Voting and inter-agent Debate--and assess their respective contributions. Through extensive experiments across seven NLP benchmarks, we find that Majority Voting alone accounts for most of the performance gains typically attributed to MAD. To explain this, we propose a theoretical framework that models debate as a stochastic process. We prove that it induces a martingale over agents' belief trajectories, implying that debate alone does not improve expected correctness. Guided by these insights, we demonstrate that targeted interventions, by biasing the belief update toward correction, can meaningfully enhance debate effectiveness. Overall, our findings suggest that while MAD has potential, simple ensembling methods remain strong and more reliable alternatives in many practical settings. Code is released in https://github.com/deeplearning-wisc/debate-or-vote.
ParlAI Vote: A Web Platform for Analyzing Gender and Political Bias in Large Language Models
Purdue University（普渡大学）；Johns Hopkins University（约翰霍普金斯大学）；Vanderbilt University（范德比大学）；University of Toronto（多伦多大学）；The University of Melbourne（墨尔本大学）
Debate to Align: Reliable Entity Alignment through Two-Stage Multi-Agent Debate
Hubei Provincial Key Laboratory of Artificial Intelligence and Smart Learning, Central China Normal University, Wuhan, China（湖北人工智能与智能学习省级重点实验室，中央财经大学，武汉，中国）；School of Computer Science, Central China Normal University, Wuhan, China（中央财经大学计算机科学学院，武汉，中国）；National Language Resources Monitoring and Research Center for Network Media, Central China Normal University, Wuhan, China（网络媒体语言资源监测与研究中心，中央财经大学，武汉，中国）；College of Computer Science, Inner Mongolia University, Hohhot, China（内蒙古大学计算机学院，呼和浩特，中国）；National and Local Joint Engineering Research Center of Intelligent Information Processing Technology for Mongolian, Inner Mongolia University, Hohhot, China（蒙古语智能信息处理技术国家与地方联合工程研究中心，内蒙古大学，呼和浩特，中国）；Inner Mongolia Key Laboratory of Multilingual Artificial Intelligence Technology, Inner Mongolia University, Hohhot, China（内蒙古多语言人工智能技术重点实验室，内蒙古大学，呼和浩特，中国）
AgentsCourt: Building Judicial Decision-Making Agents with Court Debate Simulation and Legal Knowledge Augmentation
Institute of Automation, Chinese Academy of Sciences（中国科学院自动化研究所）；School of Artificial Intelligence, University of Chinese Academy of Sciences（中国科学院大学人工智能学院）
A Structured Debate-Mixture-of-Agents Framework for Complex Clinical Diagnostic Decision Support
College of Computer Science, Sichuan University（四川大学计算机学院）；West China Hospital, Sichuan University（四川大学华西医院）；Med-X Center for Informatics, Sichuan University（四川大学华西医学信息学中心）
Benchmarking Gender and Political Bias in Large Language Models
School of Computing & Information Systems, The University of Melbourne（墨尔本大学计算与信息系统学院）；Mohamed bin Zayed University of Artificial Intelligence（穆罕默德·本·扎耶德人工智能大学）
Red and blue language: Word choices in the Trump & Harris 2024 presidential debate
Institute for Information and Language Processing (CIS), LMU Munich（慕尼黑大学信息与语言处理研究所）；University Bologna（博洛尼亚大学）；Munich Center of Machine Learning (MCML)（慕尼黑机器学习中心）
Through the Lens of Split Vote: Exploring Disagreement, Difficulty and Calibration in Legal Case Outcome Classification
Technical University of Munich（慕尼黑工业大学）；Graduate Institute of International and Development Studies（国际与发展研究研究生院）；IT University of Copenhagen（哥本哈根IT大学）；LMU Munich（慕尼黑大学）；Munich Center for Machine Learning (MCML)（慕尼黑机器学习中心）；ELTEMATE
