SATURN enables scalable, verifiable, and curriculum-controlled reinforcement learning that enhances the reasoning capability of LLMs, and generalizes to math and programming tasks.