1 lesson tagged rlvr: free, quiz-checked micro-lessons.
How Group Relative Policy Optimization turns unit tests and theorem provers into training signals — the technique behind DeepSeek-R1's leap in math and code reasoning without a reward model.
