2025 · Conference paper
UIOrchestra: Generating High-Fidelity Code from UI Designs with a Multi-agent System
UIOrchestra converts static mobile UI designs into React/CSS through layout-description, code-generation and rendered-image comparison agents. Its APPUI benchmark contains 1,101 human-refined code/rendering pairs based on screenshots from 165 apps across 12 categories. The reference code is reconstructed for evaluation, rather than recovered original application source.
Scope and limitations
Tests visual, component and code similarity, not native-package execution, interaction, backend behavior or application equivalence. APPUI references are rendered reconstructions of iOS screenshots; unavailable images/icons use placeholders. A human study covers 200 samples with four annotators. The difference-analysis prompt ignores font-size/typeface differences and placeholder image content. Model snapshots, rendering dependencies and a reproducible runtime setup were not verified; experiments were not rerun.
Why it fits this collection
Adds a mobile-oriented screenshot reconstruction method and benchmark with explicit human-built reference data; also evaluates on Design2Code-hard webpages.
Publication and review details
- Venue
- Findings of EMNLP 2025
- Status
- Conference paper
- Publication / posting date
- November 2025
- Date source
- The official ACL Anthology citation gives November 2025. Crossref records only the year; the November 4–9 conference dates printed in the PDF are not treated as a publication day.
- DOI
- 10.18653/v1/2025.findings-emnlp.150
Reviewed the final 14-page proceedings PDF. Its eleven-author order is retained, including Yanping Wang where ACL/Crossref metadata renders Wangyanping; Wang Yixuan is preserved as printed. No arXiv version was verified. The paper promises a code/data release, but the reviewed coauthor page has disabled code/project placeholders and no attributable release or artifact license was verified. Code/project URLs therefore remain unset. The paper is covered by ACL’s CC BY 4.0 policy; this does not establish licensing for an unreleased dataset or depicted apps.
Sources reviewed
Source review only; experiments were not rerun for this collection. Cite the original authors using the linked paper or BibTeX.