🌾 OAT: A research-friendly framework for LLM online alignment, including reinforcement learning, preference learning, etc. - View it on GitHub