A Theoretical Perspective: How to Prevent Model Collapse in Self-consuming Training Loops.

Shi Fu, Yingjie Wang, Yuzhu Chen, Xinmei Tian, Dacheng Tao

Browse the full ICLR paper archive.