AI-generated data contamination erodes pathological variability and diagnostic reliability
Published in arXiv preprint, 2026
This preprint investigates how AI-generated data contamination erodes pathological variability and diagnostic reliability in medical AI systems. The work connects to ongoing research on model collapse in multi-modal medical generative models at NUS CogAI4Sci.
Recommended citation: H. He, S. Xiang, Y. Zhang, Y. Zhu, J. Zhang, H. Deng, E. Alsentzer, Q. Chen, K.-H. Yu, A. Marmenshall, T. Chen, S. Anumasa, D. Ebner, D. Ho, K. Y. Ngiam, C.-Y. Cheng, and D. Liu. (2026). "AI-generated data contamination erodes pathological variability and diagnostic reliability." arXiv:2601.12946.
Download Paper