Predicting Future Actions of Reinforcement Learning Agents

Furthermore, using internal plans proves more robust to model quality compared to simulation-based approaches when predicting actions, while the results for event prediction are more mixed.