Bar-by-Bar Feedback: How Dense Rewards Teach AI to Reason

od Intellectually Curious