AI 中文总结
该研究以《狼人杀》为框架，分析LLM多智能体的目标不一致问题，发现其会破坏对抗环境结果且在公开行为中难察觉，凸显需缓解策略。
AI 中文摘要
基于大型语言模型（LLM）的多智能体系统越来越多地部署在混合动机环境中，在此类环境中，智能体因目标冲突或隐藏而在不对称信息和策略欺骗下运行，与集体目标的不一致成为核心问题。我们提出一种使用社交推理游戏《狼人杀》评估目标不一致的新框架，在保留单个智能体指定角色的同时修改其目标。针对四个不同模型系列及规模的LLM、四个玩家角色和三种目标表述，我们对智能体的内部推理和公开廉价谈话行为（即不直接影响智能体效用的无成本、非约束性交流）进行双重分析，并补充游戏结果分析。结果显示，目标不一致会破坏固有对抗环境中的结果，且该效应因不对称信息和专门角色而加剧；受损智能体始终会形成不同的、依赖目标的推理策略，但这些调整在其公开行为中基本不可见。更广泛而言，我们的发现表明，即使是细微的目标不一致也会深刻影响集体决策，凸显了基于LLM的多智能体系统需要有效缓解策略。
英文摘要
Large Language Models (LLMs)-powered multi-agent systems are increasingly deployed in mixed-motive environments, where agents operate under asymmetric information and strategic deception due to conflicting or hidden objectives. In these settings, misalignment with collective goals becomes a central concern. We propose a novel framework for evaluating objective misalignment using the social deduction game Werewolf, modifying the objective of a single agent while preserving its assigned role. Across LLMs from four different model families and sizes, four player roles, and three objective formulations, we introduce a dual analysis of the agents' internal reasoning and their public cheap-talk behavior (i.e costless, non-binding communication that does not directly affect the agents' utilities), complemented by an analysis of game outcomes. Our results show that objective misalignment undermines outcomes in inherently adversarial environments, an effect exacerbated by asymmetric information and specialized roles. While compromised agents consistently develop distinct objective-dependent reasoning strategies, these adaptations remain largely invisible in their public behavior. More broadly, our findings suggest that even subtle objective misalignment can profoundly affect collective decision-making, highlighting the need for effective mitigation strategies for LLM-based multi-agent systems.
CommentsAccepted at AIWILD@ICLR 2026
