REBEL: A Regularization-Based Solution for Reward Overoptimization in Robotic Reinforcement Learning from Human Feedback

arXiv:2312.14436v2 Announce Type: replace-cross
Abstract: The effectiveness of reinforcement learning (RL) agents in continuous control robotics tasks is heavily dependent on the design of the underlying reward function. However, a misalignment between the reward function and user intentions, values, or social norms can be catastrophic in the real world. Current methods to mitigate this misalignment work by learning reward functions from human preferences; however, they inadvertently introduce a risk of reward overoptimization. In this work, we address this challenge by …

abstract agents arxiv continuous control cs.lg cs.ro design feedback function however human human feedback regularization reinforcement reinforcement learning robotic robotics social solution tasks type values