Choices are More Important than Efforts: LLM Enables Efficient Multi-Agent Exploration

TL;DR

This paper presents LEMAE, a systematic approach leveraging large language models to guide multi-agent exploration efficiently by focusing on key states, significantly improving performance and reducing redundant efforts in complex reinforcement learning tasks.

Contribution

The paper introduces LEMAE, a novel framework that uses LLMs to identify key states and guide exploration, reducing redundancy and accelerating learning in multi-agent reinforcement learning.

Findings

LEMAE outperforms state-of-the-art methods on benchmarks like SMAC and MPE.

Achieves up to 10x faster exploration in certain scenarios.

Effectively reduces redundant exploration efforts.

Abstract

With expansive state-action spaces, efficient multi-agent exploration remains a longstanding challenge in reinforcement learning. Although pursuing novelty, diversity, or uncertainty attracts increasing attention, redundant efforts brought by exploration without proper guidance choices poses a practical issue for the community. This paper introduces a systematic approach, termed LEMAE, choosing to channel informative task-relevant guidance from a knowledgeable Large Language Model (LLM) for Efficient Multi-Agent Exploration. Specifically, we ground linguistic knowledge from LLM into symbolic key states, that are critical for task fulfillment, in a discriminative manner at low LLM inference costs. To unleash the power of key states, we design Subspace-based Hindsight Intrinsic Reward (SHIR) to guide agents toward key states by increasing reward density. Additionally, we build the Key State Memory Tree (KSMT) to track transitions between key states in a specific task for organized exploration. Benefiting from diminishing redundant explorations, LEMAE outperforms existing SOTA approaches on the challenging benchmarks (e.g., SMAC and MPE) by a large margin, achieving a 10x acceleration in certain scenarios.