Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2406.19644 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866909234004754432 |
|---|---|
| author | Shen, Zichao Zhu, Tianchen Sun, Qingyun Gao, Shiqi Li, Jianxin |
| author_facet | Shen, Zichao Zhu, Tianchen Sun, Qingyun Gao, Shiqi Li, Jianxin |
| contents | Reinforcement learning (RL) faces challenges in evaluating policy trajectories within intricate game tasks due to the difficulty in designing comprehensive and precise reward functions. This inherent difficulty curtails the broader application of RL within game environments characterized by diverse constraints. Preference-based reinforcement learning (PbRL) presents a pioneering framework that capitalizes on human preferences as pivotal reward signals, thereby circumventing the need for meticulous reward engineering. However, obtaining preference data from human experts is costly and inefficient, especially under conditions marked by complex constraints. To tackle this challenge, we propose a LLM-enabled automatic preference generation framework named LLM4PG , which harnesses the capabilities of large language models (LLMs) to abstract trajectories, rank preferences, and reconstruct reward functions to optimize conditioned policies. Experiments on tasks with complex language constraints demonstrated the effectiveness of our LLM-enabled reward functions, accelerating RL convergence and overcoming stagnation caused by slow or absent progress under original reward structures. This approach mitigates the reliance on specialized human knowledge and demonstrates the potential of LLMs to enhance RL's effectiveness in complex environments in the wild. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2406_19644 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Beyond Human Preferences: Exploring Reinforcement Learning Trajectory Evaluation and Improvement through LLMs Shen, Zichao Zhu, Tianchen Sun, Qingyun Gao, Shiqi Li, Jianxin Artificial Intelligence Reinforcement learning (RL) faces challenges in evaluating policy trajectories within intricate game tasks due to the difficulty in designing comprehensive and precise reward functions. This inherent difficulty curtails the broader application of RL within game environments characterized by diverse constraints. Preference-based reinforcement learning (PbRL) presents a pioneering framework that capitalizes on human preferences as pivotal reward signals, thereby circumventing the need for meticulous reward engineering. However, obtaining preference data from human experts is costly and inefficient, especially under conditions marked by complex constraints. To tackle this challenge, we propose a LLM-enabled automatic preference generation framework named LLM4PG , which harnesses the capabilities of large language models (LLMs) to abstract trajectories, rank preferences, and reconstruct reward functions to optimize conditioned policies. Experiments on tasks with complex language constraints demonstrated the effectiveness of our LLM-enabled reward functions, accelerating RL convergence and overcoming stagnation caused by slow or absent progress under original reward structures. This approach mitigates the reliance on specialized human knowledge and demonstrates the potential of LLMs to enhance RL's effectiveness in complex environments in the wild. |
| title | Beyond Human Preferences: Exploring Reinforcement Learning Trajectory Evaluation and Improvement through LLMs |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2406.19644 |