On The Fragility of Learned Reward Functions
This work addresses a critical issue in reinforcement learning for AI safety, highlighting a fragility in reward learning that could lead to misaligned behaviors, though it is incremental in focusing on evaluation methods rather than proposing a new solution.
The paper tackles the problem of learned reward functions failing to train new policies from scratch in preference-based reward learning, demonstrating through experiments in tabular and continuous control environments that relearning failures are sensitive to reward model design and dataset composition.
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.