Prm800k

PRM800K is a prominent, open-source dataset consisting of 800,000 step-by-step process supervision labels built by OpenAI. It is explicitly engineered to train large language models to accurately track and solve complex multi-step mathematical reasoning logic.