VFIG enables high-fidelity conversion of rasterized technical figures into editable, scalable SVGs using a new 66K-pair dataset.
March 26, 2026
Original Paper
VFIG: Vectorizing Complex Figures in SVG with Vision-Language Models
arXiv · 2603.24575
The Takeaway
Reconstructing vector source files from PNG/JPEGs is a major pain point in technical design and paper writing. The release of VFIG-DATA and the associated coarse-to-fine RL training curriculum provides a robust tool for democratizing figure editing and high-quality diagram generation.
From the abstract
Scalable Vector Graphics (SVG) are an essential format for technical illustration and digital design, offering precise resolution independence and flexible semantic editability. In practice, however, original vector source files are frequently lost or inaccessible, leaving only "flat" rasterized versions (e.g., PNG or JPEG) that are difficult to modify or scale. Manually reconstructing these figures is a prohibitively labor-intensive process, requiring specialized expertise to recover the origin