On The Fragility of Learned Reward Functions

TL;DR

This paper investigates the fragility of learned reward functions in preference-based reward learning, highlighting that policy performance alone can mask underlying reward misalignments and advocating for more rigorous retraining evaluations.

Contribution

It demonstrates the causes of reward relearning failures and shows their sensitivity to reward model design and dataset composition, emphasizing the need for better evaluation methods.

Findings

Relearning failures can occur even when policies perform well.

Reward model design and dataset composition influence relearning success.

Current evaluations may overlook reward misalignment issues.

Abstract

Reward functions are notoriously difficult to specify, especially for tasks with complex goals. Reward learning approaches attempt to infer reward functions from human feedback and preferences. Prior works on reward learning have mainly focused on the performance of policies trained alongside the reward function. This practice, however, may fail to detect learned rewards that are not capable of training new policies from scratch and thus do not capture the intended behavior. Our work focuses on demonstrating and studying the causes of these relearning failures in the domain of preference-based reward learning. We demonstrate with experiments in tabular and continuous control environments that the severity of relearning failures can be sensitive to changes in reward model design and the trajectory dataset composition. Based on our findings, we emphasize the need for more retraining-based evaluations in the literature.