Reward-Aware Proto-Representations in Reinforcement Learning | Read Paper on Bytez