Learn as Individuals, Evolve as a Team: Multi-agent LLMs Adaptation in Embodied Environments

2025-06-08

AI Summary

Existing LLM-based multi-agent planning methods exhibit limited adaptability in embodied environments and struggle with dynamic collaborative tasks. To address this, we propose the LIET paradigm, which establishes a dual-path adaptive mechanism: “individual learning” enables embodied perception and real-time decision-making via local utility functions, while “team evolution” achieves communication coordination and policy consensus through an iteratively updated shared collaboration knowledge list. LIET integrates exploratory data-driven supervised learning, runtime utility querying, and iterative consensus updates over the knowledge list, enabling end-to-end cooperative optimization within LLaMA/GPT-4o multi-agent frameworks. Evaluated on the Communicative Watch-And-Help and ThreeD-World Multi-Agent Transport benchmarks, LIET significantly outperforms state-of-the-art methods, demonstrating superior robustness and strong cross-scenario generalization capability.

Technology Category

• Multiagent Systems: Multiagent Planning
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Search and Optimization: Learning to Search

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Large language models (LLMs) possess extensive knowledge bases and strong reasoning capabilities, making them promising tools for complex, multi-agent planning in embodied environments. However, despite LLMs' advanced abilities and the sophisticated modular design of agentic methods, existing LLM-based planning algorithms remain limited by weak adaptation capabilities to multi-agent embodied scenarios. We address this limitation by introducing a framework that enables LLM agents to learn and evolve both before and during test time, equipping them with environment-relevant knowledge for better planning and enhanced communication for improved cooperation. Inspired by centralized training with decentralized execution in multi-agent reinforcement learning, we propose a extit{Learn as Individuals, Evolve as a Team (LIET)} paradigm for multi-agent LLMs adaptation. At the individual level, LLM agents learn a local utility function from exploratory datasets to better comprehend the embodied environment, which is then queried during test time to support informed decision-making. At the team level, LLM agents collaboratively and iteratively maintain and update a shared cooperation knowledge list based on new experiences, using it to guide more effective communication. By combining individual learning with team evolution, LIET enables comprehensive and flexible adaptation for LLM agents. Our experiments on Communicative Watch-And-Help and ThreeD-World Multi-Agent Transport benchmarks demonstrate that LIET, instantiated with both LLaMA and GPT-4o, outperforms existing baselines and exhibits strong cooperative planning abilities.

Problem

• Enhancing multi-agent LLM adaptation in embodied environments
• Improving planning and cooperation through individual and team learning
• Addressing weak adaptation in LLM-based multi-agent planning algorithms

Innovation

• Multi-agent LLMs learn local utility functions
• Team-level shared cooperation knowledge updating
• Combined individual learning and team evolution