Reinforcement learning and llms