How Do We Teach Reinforcement Learning Agents Human Preferences?
by Language Models (dot tech)December 3rd, 2024
Too Long; Didn't Read
ICPL enhances reinforcement learning by integrating Large Language Models with human preference feedback, refining reward functions interactively. Building on works like EUREKA, it sets new standards in RL reward design efficiency.
Table of Links
- Abstract and Introduction
- Related Work
- Problem Definition
- Method
- Experiments
- Conclusion and References
A. Appendix
A.1. Full Prompts and A.2 ICPL Details
A.6 Human-in-the-Loop Preference
3 PROBLEM DEFINITION
Our goal is to design a reward function that can be used to train reinforcement learning agents that demonstrate human-preferred behaviors. It is usually hard to design proper reward functions in reinforcement learning that induce policies that align well with human preferences.
Authors:
(1) Chao Yu, Tsinghua University;
(2) Hong Lu, Tsinghua University;
(3) Jiaxuan Gao, Tsinghua University;
(4) Qixin Tan, Tsinghua University;
(5) Xinting Yang, Tsinghua University;
(6) Yu Wang, with equal advising from Tsinghua University;
(7) Yi Wu, with equal advising from Tsinghua University and the Shanghai Qi Zhi Institute;
(8) Eugene Vinitsky, with equal advising from New York University ([email protected]).
This paper is available on arxiv under CC 4.0 license.
L O A D I N G
. . . comments & more!
. . . comments & more!
