Microsoft AI Introduces rStar2-Agent: A 14B Math Reasoning Model Trained with Agentic Reinforcement Learning to Achieve Frontier-Level Performance Deja un comentario / Por / agosto 30, 2025