Tao Li

DuPO: Enabling Reliable Self-Verification via Dual Preference Optimization