Cross-domain Random Pre-training with Prototypes for Reinforcement Learning | Read Paper on Bytez