Abstract: Large Vision-Language Models (LVLMs) suffer from severe object hallucinations, leading them to frequently generate outputs that do not correspond to the image content, significantly reducing ...

Abstract: Existing free-energy guided No-Reference Image Quality Assessment (NR-IQA) methods continue to face challenges in effectively restoring complexly distorted images. The features guiding the ...