MS student, University of Science and Technology of China
1 paper at NeurIPS 2025
This paper proves that self-verification prevents model collapse in recursive training without relying on real data.
