Researchers developed a vision-language model pipeline that generates dense, grounded captions for comic panels, improving accessibility and understanding for visually impaired individuals. Their ...