RCG-for-reinforcement-learning