On the Importance of Exploration for Generalization in Reinforcement Learning

We hypothesize that the agent's exploration strategy plays a key role in its ability to generalize to new environments.