August 16, 2025

Improving Sample Efficiency of Reinforcement Learning With Background Knowledge From Large Language Models.

Key Points

Significant improvements in sample efficiency were achieved, indicating better learning outcomes for reinforcement learning tasks.
With up to three variants, background knowledge was extracted from the environment for improved policy learning.
Using potential functions, the approach enables effective reward shaping while maintaining task optimality for various scenarios.
Findings emphasize the importance of leveraging large language models, positioning them as valuable tools in reinforcement learning frameworks.

Abstract

Low sample efficiency is an enduring challenge of reinforcement learning (RL). With the advent of versatile large language models (LLMs), recent works impart common-sense knowledge to accelerate policy learning for RL processes. However, we note that such guidance is often tailored for one specific task but loses generalizability. In this article, we introduce a framework that harnesses LLMs to extract background knowledge of an environment, which contains general understandings of the entire environment, making various downstream RL tasks benefit from one-time knowledge representation. We ground LLMs by feeding a few precollected experiences and requesting them to delineate background knowledge of the environment. Afterward, we represent the output knowledge as potential functions for potential-based reward shaping, which has a good property for maintaining policy optimality from task rewards. We instantiate three variants to prompt LLMs for background knowledge, including writing code, annotating pReferences, and assigning goals. Our experiments show that these methods achieve significant sample efficiency improvements in a spectrum of downstream tasks from Minigrid and Crafter domains.

Mark Helpful

Bookmark

Relay