{
  "added_at": "2026-10-17",
  "arxiv_id": "2507.22827",
  "arxiv_version_reviewed": "v2",
  "authors": [
    "Yilei Jiang",
    "Yaozhi Zheng",
    "Yuxuan Wan",
    "Jiaming Han",
    "Qunzhong Wang",
    "Michael R. Lyu",
    "Xiangyu Yue"
  ],
  "bibtex_url": "https://awesomerea.org/papers/citations/paper-screencoder.bib",
  "code_url": "https://github.com/leigest519/ScreenCoder",
  "date_basis": "The official EMNLP 2026 detailed program lists a main-conference poster, verifying acceptance; no proceedings publication date is asserted.",
  "date_basis_zh": "EMNLP 2026 官方详细日程列出了该论文的主会海报报告，可确认已被接收；不推定论文集的出版日期。",
  "date_precision": "year",
  "description_en": "Splits screenshot-to-code generation among grounding, planning and generation agents that label UI regions, build a hierarchical layout tree and write HTML/CSS. The pipeline also produces the Screen-10K training pairs used for supervised and reinforcement fine-tuning, and the paper introduces the 1,000-page ScreenBench.",
  "description_zh": "将截图到代码的生成分配给定位、规划和生成三个智能体，分别标注界面区域、构建层级布局树并编写 HTML/CSS。该流程还生成了用于监督微调和强化学习的 Screen-10K 训练数据；论文另提出包含 1,000 个网页的 ScreenBench。",
  "doi": null,
  "entry_url": "https://awesomerea.org/papers/paper-screencoder/",
  "entry_url_zh": "https://awesomerea.org/zh/papers/paper-screencoder/",
  "evidence_urls": [
    "https://arxiv.org/abs/2507.22827",
    "https://2026.emnlp.org/program/",
    "https://docs.google.com/spreadsheets/d/1aXGTy_7Xeh-OXIs3iJSbDpUsZ6YhbfA76b17kBZH0Yk/edit?gid=943432490#gid=943432490",
    "https://openreview.net/forum?id=29VraifZkB",
    "https://github.com/leigest519/ScreenCoder/blob/22816fdfb4ee5982973f9a09d67498a64823e65d/README.md",
    "https://github.com/leigest519/ScreenCoder/blob/22816fdfb4ee5982973f9a09d67498a64823e65d/LICENSE",
    "https://jimmyzhengyz.github.io/ScreenCoder-Project-Page/",
    "https://huggingface.co/datasets/Leigest/ScreenCoder/blob/b0e49d2a5d937e93f712d4c281d4036ec2652a30/README.md"
  ],
  "first_arxiv_date": "2025-07-30",
  "id": "paper-screencoder",
  "json_url": "https://awesomerea.org/papers/paper-screencoder/record.json",
  "kind": "research",
  "markdown_url": "https://awesomerea.org/papers/paper-screencoder/index.md",
  "markdown_url_zh": "https://awesomerea.org/zh/papers/paper-screencoder/index.md",
  "metadata_notes": "The official EMNLP 2026 detailed program lists paper 7281-MAIN as a poster on 2026-10-26; this records acceptance, not proceedings publication. The arXiv v2 comment 'ScreenCoder-v2' is a version label, not a venue, and an earlier ICLR 2026 submission was withdrawn. The abstract-linked repository is Apache-2.0 at the pinned commit and bundles UIED, LLaMA-Factory and VLM-R1 code with their own license files. ScreenBench is published on Hugging Face under Apache-2.0.",
  "metadata_notes_zh": "EMNLP 2026 官方详细日程将论文 7281-MAIN 列为 2026-10-26 的海报报告；本条记录接收状态，不视为论文集已出版。arXiv v2 备注“ScreenCoder-v2”是版本标签而非会议名称；此前提交 ICLR 2026 的版本已撤回。摘要所链接的仓库在固定提交中采用 Apache-2.0 许可，并内含带有各自许可证文件的 UIED、LLaMA-Factory 与 VLM-R1 代码。ScreenBench 以 Apache-2.0 许可发布于 Hugging Face。",
  "pdf_url": "https://arxiv.org/pdf/2507.22827v2",
  "project_url": "https://jimmyzhengyz.github.io/ScreenCoder-Project-Page/",
  "publication_date": null,
  "publication_status": "conference-accepted",
  "record_type": "paper",
  "runtime_tested": false,
  "scope_limitations_en": "Output is static HTML/CSS layout; interaction, JavaScript behavior and backend logic are not reconstructed. Results are visual-similarity scores (CLIP and OCR-matched text blocks) on ScreenBench and Design2Code, which do not establish code-level or functional equivalence. The released pipeline calls a hosted model API, Doubao by default. The reviewed text is arXiv v2; the accepted EMNLP version may differ.",
  "scope_limitations_zh": "输出为静态 HTML/CSS 布局，不重建交互、JavaScript 行为或后端逻辑。结果是 ScreenBench 与 Design2Code 上的视觉相似度分数（CLIP 及经 OCR 匹配的文本块），不能证明代码层面或功能上的等价。公开流程需调用托管模型 API，默认使用豆包（Doubao）。审阅文本为 arXiv v2，EMNLP 接收版本可能有所不同。",
  "targets": [
    "website"
  ],
  "title": "ScreenCoder: Advancing Visual-to-Code Generation for Front-End Automation via Modular Multimodal Agents",
  "title_zh": "ScreenCoder：以模块化多模态智能体推进面向前端自动化的视觉到代码生成",
  "url": "https://arxiv.org/abs/2507.22827",
  "venue": "EMNLP 2026",
  "verification": "source-reviewed",
  "verified_at": "2026-10-17",
  "why_fits_en": "Represents staged, agent-based screenshot-to-code reconstruction for websites, separating visual grounding and layout planning from code synthesis.",
  "why_fits_zh": "代表分阶段、基于智能体的网站截图到代码重建方法，将视觉定位和布局规划与代码生成分开处理。",
  "year": 2026
}