MulCLIP: A Multi-level Alignment Framework for Enhancing Fine-grained Long-context CLIP

Chau Truong, Hieu Ta Quang, Dung D. Le

Vision-language models like CLIP show impressive ability to align images and text, but their training on short, concise captions makes them struggle with lengthy, detailed descript…