LLawCo: Learning Laws of Cooperation for Modeling Embodied Multi-Agent Behavior

2026-06-26

AI Summary

This work addresses the inefficiency in collaborative embodied multi-agent systems operating in decentralized, partially observable environments, where inconsistent behaviors hinder effective coordination. To tackle this challenge, the authors propose LLawCo, a novel framework that extracts interpretable high-level cooperation rules—such as “communicate when necessary” and “wait for partners”—by reflecting on failed interaction episodes. These rules are explicitly injected into the chain-of-thought reasoning process of large language models for the first time, and alignment with task objectives and other agents is achieved through supervised fine-tuning. Evaluated on the PARTNR-Dialog and TDW-MAT benchmarks, LLawCo improves task success rates by 4.5% and 6.8%, respectively, over the current best open-source methods, demonstrating significantly enhanced collaboration efficiency and generalization capability.

Abstract

Embodied agents operating in decentralized and partially observable environments have attracted growing attention in recent years. However, existing large language model (LLM)-based agents often exhibit behaviors that are misaligned with their partners or inconsistent with the environment state, leading to inefficient cooperation and poor task success. To address this challenge, we propose a novel framework, Learning Laws of Cooperation (LLawCo), that enables embodied agents to autonomously align with both their partners and task objectives. Our framework allows agents to reflect on past failures to extract misaligned behavioral patterns, which are used to derive high-level behavioral laws, such as "Talk when necessary" and "Wait for partner." These laws are explicitly incorporated into the agents' chains of thought via supervised fine-tuning, aligning their reasoning with task requirements and the behavior of other agents. To evaluate our approach, we introduce PARTNR-Dialog, a large-scale multi-agent communicative and cooperative planning benchmark built on the PARTNR environment. Experiments on existing tasks and our new benchmark demonstrate significant improvements in cooperative efficiency and task success rates. Across four backbone LLMs, our method achieves average success rate improvements of 4.5% on the PARTNR-Dialog benchmark and 6.8% on the TDW-MAT benchmark over state-of-the-art open-source communicative agent frameworks. See the LLawCo project page for details: https://www.merl.com/research/highlights/LLawCo

Problem

• embodied agents
• multi-agent cooperation
• partial observability
• behavioral alignment
• task success

Innovation

• LLawCo
• embodied multi-agent cooperation
• behavioral laws
• chain-of-thought alignment
• supervised fine-tuning