Disentangling Transfer in Continual Reinforcement Learning Maciej Wołczyk Faculty of Mathematics and Computer Science