Exploiting Language Instructions for Interpretable and Compositional Reinforcement Learning | Read Paper on Bytez