New working paper on class abstractions in RL

Niklas Höpner has recently posted his working paper on leveraging class abstraction for commonsense reinforcement learning. It is available on the arXiv.