The original approach proposed by paper authors, reproduced and used as a reference point for evaluating AI-Scientist performance.