AI RESEARCH

CHEM: Estimating and Understanding Hallucinations in Deep Learning for Image Processing

arXiv CS.CV

ArXi:2512.09806v2 Announce Type: replace Deep learning-based methods have recently achieved significant success in image reconstruction problems. However, challenges have emerged, as these methods may generate unrealistic artifacts or hallucinations, which can interfere with analysis in safety-critical scenarios. This paper