“Prior to O1, many high-profile Deep Reinforcement Learning results were narrow in their applicability and did not achieve an impact comparable to models like GPT-4.”