Multimodal models struggle to edit infographics with proper layout reasoning
InfoEdit: Probing Global Layout Reasoning in Infographic Editing
Computer Vision and Pattern RecognitionComputation and LanguageSoftware Engineering
Summary
Editing photos with AI is getting good, but editing infographics is much harder because changing one part often means many others need adjustment too. The authors call this tricky problem "reflow." They created InfoEdit, a dataset and test system with 1,000 infographics and 4,000 editing tasks designed to see how well AI models manage these layout changes. Most current models perform poorly, showing this is a big challenge. They also found that editing with the underlying code can sometimes be as good as editing pixels directly.
What this means in practice
- •For graphic designers: Improve tools that edit infographic layouts by better handling global repositioning of elements for clear information flow.
- •For software developers: Develop infographic editing software that uses benchmark data to enhance AI-driven layout adjustments during edits.
Authors
Cheng Yang, Chufan Shi, Huijuan Wang, Bo Shui, Yaokang Wu, Muzi Tao, Yibo Yan, Xuezhe Ma, Taylor Berg-Kirkpatrick
Abstract
Multimodal foundation models edit natural photographs at production quality, yet the same models struggle with structured visual content such as infographics. Unlike photographs, infographics encode information through logical relations; editing one element often requires surrounding elements to be adapted. We refer to this global layout reasoning capability as reflow. Existing image-editing benchmarks neither provide a dedicated setting for structured visual content nor evaluate the reflow capability. We introduce InfoEdit, a novel benchmark of 1,000 infographics across eight logical-relation families, paired with 4,000 editing instructions across four editing tasks, and a reflow-aware evaluation protocol. Across eight frontier editors, only GPT-Image-2 clears 60% average success rate; most models fall below 7%, and no editor exceeds 36% on the Swap-Block task even with perfect target localization. We further show that code-level editing can match the strongest pixel-level editor, revealing complementary strengths across tasks. InfoEdit identifies reflow as a central challenge in structured visual content editing and provides a diagnostic benchmark to facilitate future progress.