1 paper
Zixiong Yu, Jun Rao, Guhan Chen +5
Synthesizing high-quality mathematical reasoning data without human priors remains a significant challenge. Current approaches typically rely on seed data mutation or simple prompt…