多智能体系统的心智模型框架

AI 摘要

该论文提出「心智模型赋能智能体」框架，为智能体配备关于对手的潜在心智模型，使其能从观察历史中推断隐藏的信念、意图和可能反应，并据此指导行动选择。方法学习一种摊销递归的心智理论表示，包含一阶和二阶心理状态结构，并与信念条件奖励模型联合训练，最终在信念感知信号下学习策略。在纯语言和多模态基准上，显式心智状态建模均持续提升交互质量和心智理论表现，表明结构化伙伴建模是通用多智能体系统的有效归纳偏置。代码已公开。

正文节选

Mental-Models for Multi-Agent Systems Abstract Large foundation models have accelerated progress toward general-purpose agents that interact with humans and other agents through language and multimodal signals. However, robust multi-agent decision-making requires reasoning about what other agents know, intend, and are likely to do under partial observability. Current agentic systems often operate through prompt design, memory, or end-to-end behavioral shaping, but typically do not learn an expli

发布时间：2026-10-09 12:00

抓取时间：2026-10-09 12:22

来源机构：arXiv