Believed but Not Said: Measuring Hidden Belief Collapse in Multi-Agent LLM Discussions

arXiv:2610.22138v1 Announce Type: new Abstract: LLM systems increasingly rely on groups of agents that discuss, critique, and check one another, assuming that peers will catch individual errors. We test this assumption against a minimal adversary. In a ten-agent factual discussion, one or three adversarial agents repeatedly assert a plausible false answer. After each round we probe each benign agent's private answer through a separate call whose response is hidden from its peers, and we separately label its public stance toward the claim. Across six models and three factual-QA datasets, groups remain stable without an attacker: only 0.6% of benign agents finish with the false answer. A single adversarial agent increases this rate to 44.8% and three to 60.5%. Studies of conformity typically report public agreement without private change; we find the reverse. With three adversarial agents present, benign agents hold the false answer in 51.6% of exposed turns but publicly endorse it in only 19.7%. Much of the influence therefore remains invisible in the discussion, and the false answers persist when agents are asked for their own views. Agents that retain the true answer often fail to defend it; once adopted, the false answer spreads among benign agents, and baseline knowledge does not predict robustness.

Related

• A Study of Belief Revision Postulates in Multi-Agent Systems (Extended Version)
• The Arbiter Agent: Continually Monitoring Multi-Agent Conversations to Detect Emergent Misalignment
• Even More Deception: Objective Misalignment in Mixed-Motive LLM Multi-Agent Systems
• When Upstream Messages Override Correct Answers: A Controlled Study of Multi-Agent LLM Collaboration

Source: arXiv cs.AI | 2026-10-16