Multi-agent discussion gains less when dissent is withheld
Abstract
Multi-agent systems of LLMs add discussion to majority voting and are therefore expected to be more capable. However, empirical reports conflict on whether discussion improves accuracy or leads to an incorrect consensus. Here, we introduce a parsimonious model that explains when discussion improves accuracy and when it ends in an incorrect consensus, built from four behaviors repeatedly observed in LLM agents: (1) withholding dissent, (2) internalizing a stated answer, (3) reconsidering after seeing dissent, and (4) correcting toward the correct answer. The model shows that discussion can overturn an incorrect initial majority only when the withholding rate c is below a critical rate c* = γ/(γ+ a), set by the net correction rate γ and the internalization rate a. We estimate these rates from conversation logs with a Bayesian method and place LLM teams relative to c*. As the model predicts, the gain from discussion shrinks as withholding rises, across LLMs and on a hidden profile benchmark, HiddenBench, and MedEInst. Instructing agents not to withhold dissent increases this gain. Turning reasoning off also increases the gain, because reasoning raises the internalization rate a and keeps agents from reconsidering a minority answer. These findings reconcile the conflicting reports and identify when discussion outperforms majority voting.
Related
- Dissent enhances the credibility of groups of Bayesian conformist decision-makers
- Certifying information flow in multi-agent behaviours
- Stackelberg Game with Multi-Agent Transformer for ISRS-Aware Resource Re-optimization in Multi-Band Optical Networks
- Teleportation Game: Quantum Teleportation in Multi-Agent Systems for Interactive Music
- Hierarchical Multi-Agent Reasoning Systems for Knowledge-Grounded Material Substitution
BibTeX
@misc{mansuri2026multiagent,
title = {{Multi-agent discussion gains less when dissent is withheld}},
author = {Chand Sahil Mansuri and Xin Wang and Mengying Li and Bryan Acton and Rory Eckardt and Dhaval Patel and Sadamori Kojaku},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.38324},
eprint = {2609.38324},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.38324},
}
RIS
TY - UNPB
TI - Multi-agent discussion gains less when dissent is withheld
AU - Chand Sahil Mansuri
AU - Xin Wang
AU - Mengying Li
AU - Bryan Acton
AU - Rory Eckardt
AU - Dhaval Patel
AU - Sadamori Kojaku
PY - 2026
DA - 2026/09/29
PB - arXiv
DO - 10.48550/arxiv.2609.38324
UR - https://arxiv.org/abs/2609.38324
AB - Multi-agent systems of LLMs add discussion to majority voting and are therefore expected to be more capable. However, empirical reports conflict on whether discussion improves accuracy or leads to an incorrect consensus. Here, we introduce a parsimonious model that explains when discussion improves accuracy and when it ends in an incorrect consensus, built from four behaviors repeatedly observed in LLM agents: (1) withholding dissent, (2) internalizing a stated answer, (3) reconsidering after seeing dissent, and (4) correcting toward the correct answer. The model shows that discussion can overturn an incorrect initial majority only when the withholding rate c is below a critical rate c^* = γ/(γ+ a), set by the net correction rate γ and the internalization rate a. We estimate these rates from conversation logs with a Bayesian method and place LLM teams relative to c^*. As the model predicts, the gain from discussion shrinks as withholding rises, across LLMs and on a hidden profile benchmark, HiddenBench, and MedEInst. Instructing agents not to withhold dissent increases this gain. Turning reasoning off also increases the gain, because reasoning raises the internalization rate a and keeps agents from reconsidering a minority answer. These findings reconcile the conflicting reports and identify when discussion outperforms majority voting.
ER -