The Value of Variance: Mitigating Debate Collapse in Multi-Agent Systems via Uncertainty-Driven Policy Optimization

Luoxi Tang, Yuqiao Meng, Joseph Costa, Yingxue Zhang, Muchao Ye, Zhaohan Xi
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:118988-119005, 2026.

Abstract

Multi-agent debate (MAD) systems improve LLM reasoning through iterative deliberation, but remain vulnerable to debate collapse, a failure type where final agent decisions are compromised on erroneous reasoning. Existing methods lack principled mechanisms to detect or prevent such failures. To address this gap, we first propose a hierarchical metric that quantifies behavioral uncertainty at three levels: intra-agent (individual reasoning uncertainty), inter-agent (interactive uncertainty), and system-level (output uncertainty). Empirical analysis across several benchmarks reveals that our proposed uncertainty quantification reliably indicates system failures, which demonstrates the validity of using them as diagnostic metrics to indicate the system failure. Subsequently, we propose a mitigation strategy by formulating an uncertainty-driven policy optimization to penalize self-contradiction, peer conflict, and low-confidence outputs in a dynamic debating environment. Experiments demonstrate that our proposed uncertainty-driven mitigation reliably calibrates the multi-agent system by consistently improving decision accuracy while reducing system disagreement.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-tang26n, title = {The Value of Variance: Mitigating Debate Collapse in Multi-Agent Systems via Uncertainty-Driven Policy Optimization}, author = {Tang, Luoxi and Meng, Yuqiao and Costa, Joseph and Zhang, Yingxue and Ye, Muchao and Xi, Zhaohan}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {118988--119005}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/tang26n/tang26n.pdf}, url = {https://proceedings.mlr.press/v306/tang26n.html}, abstract = {Multi-agent debate (MAD) systems improve LLM reasoning through iterative deliberation, but remain vulnerable to debate collapse, a failure type where final agent decisions are compromised on erroneous reasoning. Existing methods lack principled mechanisms to detect or prevent such failures. To address this gap, we first propose a hierarchical metric that quantifies behavioral uncertainty at three levels: intra-agent (individual reasoning uncertainty), inter-agent (interactive uncertainty), and system-level (output uncertainty). Empirical analysis across several benchmarks reveals that our proposed uncertainty quantification reliably indicates system failures, which demonstrates the validity of using them as diagnostic metrics to indicate the system failure. Subsequently, we propose a mitigation strategy by formulating an uncertainty-driven policy optimization to penalize self-contradiction, peer conflict, and low-confidence outputs in a dynamic debating environment. Experiments demonstrate that our proposed uncertainty-driven mitigation reliably calibrates the multi-agent system by consistently improving decision accuracy while reducing system disagreement.} }
Endnote
%0 Conference Paper %T The Value of Variance: Mitigating Debate Collapse in Multi-Agent Systems via Uncertainty-Driven Policy Optimization %A Luoxi Tang %A Yuqiao Meng %A Joseph Costa %A Yingxue Zhang %A Muchao Ye %A Zhaohan Xi %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-tang26n %I PMLR %P 118988--119005 %U https://proceedings.mlr.press/v306/tang26n.html %V 306 %X Multi-agent debate (MAD) systems improve LLM reasoning through iterative deliberation, but remain vulnerable to debate collapse, a failure type where final agent decisions are compromised on erroneous reasoning. Existing methods lack principled mechanisms to detect or prevent such failures. To address this gap, we first propose a hierarchical metric that quantifies behavioral uncertainty at three levels: intra-agent (individual reasoning uncertainty), inter-agent (interactive uncertainty), and system-level (output uncertainty). Empirical analysis across several benchmarks reveals that our proposed uncertainty quantification reliably indicates system failures, which demonstrates the validity of using them as diagnostic metrics to indicate the system failure. Subsequently, we propose a mitigation strategy by formulating an uncertainty-driven policy optimization to penalize self-contradiction, peer conflict, and low-confidence outputs in a dynamic debating environment. Experiments demonstrate that our proposed uncertainty-driven mitigation reliably calibrates the multi-agent system by consistently improving decision accuracy while reducing system disagreement.
APA
Tang, L., Meng, Y., Costa, J., Zhang, Y., Ye, M. & Xi, Z.. (2026). The Value of Variance: Mitigating Debate Collapse in Multi-Agent Systems via Uncertainty-Driven Policy Optimization. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:118988-119005 Available from https://proceedings.mlr.press/v306/tang26n.html.

Related Material