Analysis of Faulty Reward Functions in RL
Original title: Faulty reward functions in the wild
42AI Score
The article explores algorithm failures in reinforcement learning caused by misspecified reward functions. It highlights that such errors are often surprising and counterintuitive, representing a critical risk in agent training.
Related Entities
Full Report
Reinforcement learning algorithms can break in surprising, counterintuitive ways. In this post we’ll explore one failure mode, which is where you misspecify your reward function.