Stay Focused: Problem Drift in Multi-Agent Debate

[2502.19559] Stay Focused: Problem Drift in Multi-Agent Debate Skip to main content Search Submit Donate Log in Search arXiv Press Enter to search · Advanced search Computer Science > Computation and Language arXiv:2502.19559 (cs) [Submitted on 26 Feb 2025 ( v1 ), last revised 9 Apr 2026 (this version, v3)] Title: Stay Focused: Problem Drift in Multi-Agent Debate Authors: Jonas Becker , Lars Benedikt Kaesberg , Andreas Stephan , Jan Philip Wahle , Terry Ruas , Bela Gipp View a PDF of the paper titled Stay Focused: Problem Drift in Multi-Agent Debate, by Jonas Becker and 5 other authors View PDF HTML (experimental) Abstract: Multi-agent debate - multiple instances of large language models discussing problems in turn-based interaction - has shown promise for solving knowledge and reasoning tasks. However, these methods show limitations when solving complex problems that require longer reasoning chains. We analyze how multi-agent debate drifts away from the initial problem over multiple turns, thus harming task performance. We define this phenomenon as problem drift and quantify its presence across ten tasks (i.e., three generative, three knowledge, three reasoning, and one instruction-following task). We find that generative tasks drift often due to the subjectivity of the answer space (76-89%), compared to high-complexity tasks (7-21%). To identify the reasons, eight human experts analyze 170 multi-agent debates suffering from problem drift. We find the most common issues related to this drift are the lack of progress (35% of cases), low-quality feedback (26% of cases), and a lack of clarity (25% of cases). We propose DRIFTJudge, an LLM-as-a-judge method, as a first baseline to detect problem drift. We also propose DRIFTPolicy, which mitigates 31% of problem drift cases. Our study is a step toward understanding a key limitation of multi-agent debate, highlighting why longer debates can harm task performance and how problem drift could be addressed. Comments: accepted at EACL 2026 Subjects: Computation and Language (cs.CL) ACM classes: A.1; I.2.7 Cite as: arXiv:2502.19559 [cs.CL] (or arXiv:2502.19559v3 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2502.19559 Focus to learn more arXiv-issued DOI via DataCite Submission history From: Jonas Becker [ view email ] [v1] Wed, 26 Feb 2025 20:54:51 UTC (537 KB) [v2] Wed, 21 May 2025 14:02:49 UTC (613 KB) [v3] Thu, 9 Apr 2026 08:14:16 UTC (605 KB) Full-text links: Access Paper: View a PDF of the paper titled Stay Focused: Problem Drift in Multi-Agent Debate, by Jonas Becker and 5 other authors View PDF HTML (experimental) TeX Source view license Current browse context: cs.CL < prev | next > new | recent | 2025-02 Change to browse by: cs References & Citations NASA ADS Google Scholar Semantic Scholar export BibTeX citation Loading... BibTeX formatted citation × loading... Data provided by: Bookmark Bibliographic Tools Bibliographic and Citation Tools Bibliographic Explorer Toggle Bibliographic Explorer ( What is the Explorer? ) Connected Papers Toggle Connected Papers ( What is Connected Papers? ) Litmaps Toggle Litmaps ( What is Litmaps? ) scite.ai Toggle scite Smart Citations ( What are Smart Citations? ) Code, Data, Media Code, Data and Media Associated with this Article alphaXiv Toggle alphaXiv ( What is alphaXiv? ) Links to Code Toggle CatalyzeX Code Finder for Papers ( What is CatalyzeX? ) DagsHub Toggle DagsHub ( What is DagsHub? ) GotitPub Toggle Gotit.pub ( What is GotitPub? ) Huggingface Toggle Hugging Face ( What is Huggingface? ) ScienceCast Toggle ScienceCast ( What is ScienceCast? ) Demos Demos Replicate Toggle Replicate ( What is Replicate? ) Spaces Toggle Hugging Face Spaces ( What is Spaces? ) Spaces Toggle TXYZ.AI ( What is TXYZ.AI? ) Related Papers Recommenders and Search Tools Link to Influence Flower Influence Flower ( What are Influence Flowers? ) Core recommender toggle CORE Recommender ( What is CORE? ) Author Venue Institution Topic About arXivLabs arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs . Which authors of this paper are endorsers? | Disable MathJax ( What is MathJax? ) We gratefully acknowledge support from our major funders , member institutions , , and all contributors. About · Help · Contact · Subscribe · Copyright · Privacy · Accessibility · Operational Status (opens in new tab) Major funding support from

Related documents

• Multi-Agent Debate Strategies: Survey, Taxonomy, and Challenges#411117
• Beyond Scaling: Agents Are Heading to the Edge#200516
• Beyond Symmetric Agents: Cognitive Diversity and Multi-Agent Debate in Small Language Models#1130717
• EquiMem: Calibrating Shared Memory in Multi-Agent Debate via Game-Theoretic Equilibrium#1131525
• VACS: Value-Aligned Compositional Shielding for Multi-Agent Reasoning#1062037