Yeap, self-reinforcement learning is missing in LLMs.