Paper2Web: Let's Make Your Paper Alive!
Yuhang Chen, Tianpeng Lv, Siyi Zhang, Yixiang Yin, Yao Wan, Philip S. Yu, Dongping Chen
2025-10-20
Summary
This paper focuses on making it easier to create good websites for sharing research papers, turning complex documents into interactive and easy-to-navigate online resources.
What's the problem?
Currently, automatically building these websites is difficult. Simply using AI to write the whole website, using pre-made templates, or just converting the paper to basic web code doesn't result in websites that are well-designed, easy to use, or effectively communicate the research. Also, there wasn't a good way to *measure* how well these automatically generated websites actually worked, considering things like how easy they are to navigate, how visually appealing they are, and whether people actually understand the research after using the site.
What's the solution?
The researchers created two main things: first, a dataset called Paper2Web to test and compare different website generation methods, and a way to evaluate those websites using both automated checks and human feedback. Second, they built a system called PWAgent, which uses AI to automatically convert research papers into interactive websites with a good layout and multimedia elements. PWAgent doesn't just create a website once; it keeps improving it step-by-step to make it better.
Why it matters?
This work is important because it provides a standard way to evaluate how well AI can create research websites and offers a new system that significantly improves upon existing methods. Better research websites mean more people can access and understand scientific findings, which can speed up progress in various fields.
Abstract
Academic project websites can more effectively disseminate research when they clearly present core content and enable intuitive navigation and interaction. However, current approaches such as direct Large Language Model (LLM) generation, templates, or direct HTML conversion struggle to produce layout-aware, interactive sites, and a comprehensive evaluation suite for this task has been lacking. In this paper, we introduce Paper2Web, a benchmark dataset and multi-dimensional evaluation framework for assessing academic webpage generation. It incorporates rule-based metrics like Connectivity, Completeness and human-verified LLM-as-a-Judge (covering interactivity, aesthetics, and informativeness), and PaperQuiz, which measures paper-level knowledge retention. We further present PWAgent, an autonomous pipeline that converts scientific papers into interactive and multimedia-rich academic homepages. The agent iteratively refines both content and layout through MCP tools that enhance emphasis, balance, and presentation quality. Our experiments show that PWAgent consistently outperforms end-to-end baselines like template-based webpages and arXiv/alphaXiv versions by a large margin while maintaining low cost, achieving the Pareto-front in academic webpage generation.