Weaving Context Across Images: Improving Vision-Language Models through Focus-Centric Visual Chains

Juntian Zhang, Chuanqi cheng, Yuhan Liu +3

Vision-language models (VLMs) achieve remarkable success in single-image tasks. However, real-world scenarios often involve intricate multi-image inputs, leading to a notable perfo…