LLM agents improve decisions by learning beliefs about others
EpiAgent: Epistemic Self-Play for LLM Agents
Artificial IntelligenceComputation and LanguageMultiagent Systems
Summary
When computers work together or compete, they often need to guess what others know and how they might act. The authors created a system called EpiAgent that helps large language model agents keep track of these beliefs and update them as things happen. This helps the agents make smarter choices when they don’t have all the information. They trained the system by having agents play against themselves to get better at predicting others. Tests showed their approach was better than other methods at handling such multi-agent tasks.
What this means in practice
- •For game developers: Enhance AI-controlled characters to better estimate opponents' hidden information for improved strategic gameplay.
- •For automated negotiation teams: Build agents that infer negotiation partners' unknown preferences to plan more effective offers and responses.
Authors
Xuanfa Jin, Xinyu Cui, Zhijian Ma, Xue Yan, Jian Zhao, Haifeng Zhang, Jun Wang
Abstract
Epistemic game theory explains an agent's strategic behavior via the beliefs it holds about others and how these beliefs are updated as the interactions unfold. Modern LLM-based agents, however, often lack an explicit mechanism for learning and updating such beliefs to guide decisions under incomplete information. To bridge this gap, we propose EpiAgent, a framework that equips LLM-based agents with an explicit, learnable epistemic layer, which maintains beliefs over others' private information and combines these beliefs with the agent's available information into an epistemic state for action selection. Concretely, the agent's policy is decomposed into two coupled stages: Epistemic Update, which updates these beliefs from interaction histories, and Action Planning, which selects actions conditioned on the updated epistemic state. By formulating their dependency as a bi-level optimization problem, we introduce Epistemic Self-Play, an algorithm that alternately optimizes the two stages through self-play, utilizing an intrinsic prediction reward to evaluate how well inferred beliefs predict subsequent observations. Empirical results across both mixed and competitive scenarios show that EpiAgent consistently outperforms prompting-based and RL-based baselines. Further analyses reveal the benefits of dedicated epistemic training, improvements in belief inference accuracy, and the contribution of learned beliefs to action planning. These results demonstrate the value of an explicit, learnable epistemic layer for effective multi-agent interactions.