arXiv cs.MA· Hyeong Kyu Choi, Jiatong Li, Wendi Li, Xin Eric Wang, Sharon Li·· 12 小时前AI 评分38
多智能体 LLM 会互相探索吗？MACE 框架给出答案
Do Multi-Agent LLMs Explore Each Other?
AI 导读
研究发现现代 LLM 智能体在多智能体交互中往往缺乏有效探索，表现出短视和两极化的交互模式，导致协调效果不佳、后悔值上升。为此作者提出轻量框架 MACE，通过结构化同伴选择显式促进探索，在上下文与参数多样性设置下均显著改善探索行为和下游任务表现，并从理论上证明其期望后悔值达到次线性界。
正文 · 原文
Abstract:Exploration is essential for reliable autonomy in multi-agent systems, yet it remains unclear whether large language model (LLM) agents can explore effectively when interacting with one another. We show that modern LLM agents often fail to do so, exhibiting myopic and polarized interaction patterns that lead to suboptimal coordination and increased regret. We formalize this challenge as the Multi-Agent Exploration problem, modeling it as a partially observable stochastic game (POSG), in which agents must probe peers to infer their capabilities and identify effective interaction strategies. To address this, we introduce Multi-Agent Contextual Exploration (MACE), a lightweight framework that explicitly promotes exploration through structured peer selection. Across both contextual and parametric diversity settings, exploration behavior and downstream task performance substantially improve when MACE is applied. We further provide theoretical insight that MACE achieves a sublinear expected-regret bound. Overall, our results highlight a fundamental limitation of current LLM agents and underscore the importance of explicitly guided exploration for reliable multi-agent autonomy.
| Comments: | NeurIPS 2026 IAEval Workshop (Oral Presentation) |
| Subjects: | Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2607.11250 [cs.MA] |
| (or arXiv:2607.11250v2 [cs.MA] for this version) | |
| https://doi.org/10.48550/arXiv.2607.11250 arXiv-issued DOI via DataCite |
Submission history
From: Hyeong Kyu Choi [view email]
[v1]
Mon, 13 Jul 2026 08:34:05 UTC (1,821 KB)
[v2]
Fri, 16 Oct 2026 04:07:51 UTC (1,816 KB)
来源：arXiv cs.MA · arxiv.org