Non-Markovian Reward Modelling from Trajectory Labels via Interpretable Multiple Instance Learning
This work addresses the challenge of capturing temporal dependencies in human feedback for RL, which is incremental as it extends existing reward modelling methods to a more realistic setting.
The paper tackles the problem of reward modelling for reinforcement learning by generalizing it to handle non-Markovian rewards, removing the assumption of independent step observations, and demonstrates that their novel multiple instance learning models reconstruct reward functions with high accuracy and train high-performing policies.
We generalise the problem of reward modelling (RM) for reinforcement learning (RL) to handle non-Markovian rewards. Existing work assumes that human evaluators observe each step in a trajectory independently when providing feedback on agent behaviour. In this work, we remove this assumption, extending RM to capture temporal dependencies in human assessment of trajectories. We show how RM can be approached as a multiple instance learning (MIL) problem, where trajectories are treated as bags with return labels, and steps within the trajectories are instances with unseen reward labels. We go on to develop new MIL models that are able to capture the time dependencies in labelled trajectories. We demonstrate on a range of RL tasks that our novel MIL models can reconstruct reward functions to a high level of accuracy, and can be used to train high-performing agent policies.