Large language models (LLMs) often exhibit sycophancy: agreement with user stance even when it conflicts with the model's opinion. While prior work has mostly studied this in single-agent settings, it remains underexplored in collaborative multi-agent systems. We ask whether awareness of other agents' sycophancy levels influences discussion outcomes. To investigate this, we run controlled experiments with six open-source LLMs, providing agents with peer sycophancy rankings that estimate each peer's tendency toward sycophancy. These rankings are based on scores calculated using various static (pre-discussion) and dynamic (online) strategies. We find that providing sycophancy priors reduces the influence of sycophancy-prone peers, mitigates error-cascades, and improves final discussion accuracy by an absolute 10.5%. Thus, this is a lightweight, effective way to reduce discussion sycophancy and improve downstream accuracy.
@article{arxiv.2604.02668,
title = {Too Polite to Disagree: Understanding Sycophancy Propagation in Multi-Agent Systems},
author = {Vira Kasprova and Amruta Parulekar and Abdulrahman AlRabah and Krishna Agaram and Ritwik Garg and Sagar Jha and Nimet Beyza Bozdag and Dilek Hakkani-Tur},
journal= {arXiv preprint arXiv:2604.02668},
year = {2026}
}