Reproduce “GenAI-DrawIO-Creator: A Framework for Automated Diagram Generation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Jinze Yu, Dayuan Jiang
GenAI-DrawIO-Creator turns natural-language diagram requests into draw.io XML, validates the file, and displays it in an editable editor with conversational revision. In a 10-task evaluation using Claude 3.7 through Amazon Bedrock, first attempts covered an average of 94% of requested components and relations, while 9 of 10 XML outputs were valid. These results concern one model and curated tasks.
Key insights
- General-purpose LLMs can reliably generate complex, schema-compliant structured outputs (e.g., draw.io XML) through a combination of tailored system prompts and post-generation validation.
- Multimodal LLMs demonstrate the ability to analyze visual diagram inputs and translate them into structured descriptions suitable for subsequent diagram generation.
- An interactive, iterative framework, including real-time feedback and history management, is essential for users to refine and manage AI-generated diagrams effectively.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.