AI Models Learn to Reason During Training, Halving Parameter Needs
AI models typically learn by memorizing patterns, then researchers bolt on reasoning as an afterthought. A new method called Reinforcement Pre-Training flips this approachโteaching models to think during basic training instead.