CoSyn-400K is a multimodal synthetic question answering dataset jointly released by the University of Pennsylvania and the Allen Institute for Artificial Intelligence in 2025.Scaling Text-Rich Image Understanding via Code-Guided Synthetic Multimodal Data Generation", which aims to provide high-quality, scalable synthetic data resources for multimodal model training.

The dataset contains more than 400,000 image-text question-answering pairs, covering 10 fields such as chemistry, mathematics, nutrition, and music, 9 types of text-rich images (charts, documents, math problems, tables, charts, vector graphics, music scores, circuit diagrams, and chemical structures), and 2.7 million lines of instruction tuning data (such as image type, theme, and code generation information), supporting visual question answering tasks.