A very simple GRPO implement for reproducing r1-like LLM thinking. - View it on GitHub