VisualWebInstruct: Scaling up Multimodal Instruction Data through Web Search

Yiming Jia, Jiachen Li, Xiang Yue +4

Vision-Language Models have made significant progress on many perception-focused tasks. However, their progress on reasoning-focused tasks remains limited due to the lack of high-q…