Accelerating Reinforcement Learning Algorithms Convergence using Pre-trained Large Language Models as Tutors With Advice Reusing
This work addresses the problem of long training times in reinforcement learning for developers, though it appears incremental as it builds on existing student-teacher architectures with a focus on advice reuse.
The study tackled the slow convergence of reinforcement learning algorithms in complex environments by using pre-trained large language models as tutors, finding that this approach significantly accelerated convergence across various configurations while maintaining optimal performance.
Reinforcement Learning (RL) algorithms often require long training to become useful, especially in complex environments with sparse rewards. While techniques like reward shaping and curriculum learning exist to accelerate training, these are often extremely specific and require the developer's professionalism and dedicated expertise in the problem's domain. Tackling this challenge, in this study, we explore the effectiveness of pre-trained Large Language Models (LLMs) as tutors in a student-teacher architecture with RL algorithms, hypothesizing that LLM-generated guidance allows for faster convergence. In particular, we explore the effectiveness of reusing the LLM's advice on the RL's convergence dynamics. Through an extensive empirical examination, which included 54 configurations, varying the RL algorithm (DQN, PPO, A2C), LLM tutor (Llama, Vicuna, DeepSeek), and environment (Blackjack, Snake, Connect Four), our results demonstrate that LLM tutoring significantly accelerates RL convergence while maintaining comparable optimal performance. Furthermore, the advice reuse mechanism shows a further improvement in training duration but also results in less stable convergence dynamics. Our findings suggest that LLM tutoring generally improves convergence, and its effectiveness is sensitive to the specific task, RL algorithm, and LLM model combination.