$Q\sharp$: Provably Optimal Distributional RL for LLM Post-Training

Open in new window