← Back to Computation and Language
cs.CL

Can AI turn research papers into interactive simulations?

Dasen Dai, Biao Wu, Meng Fang, Shuoqi Li, Wenhao Wang

May 30, 2026

Most AI agents reduce papers to summaries or static webpages. This work builds a system that reads a research paper and generates a fully interactive web application—one where users can adjust inputs and watch dynamic behavior unfold. The authors created a benchmark of 19 papers with expert-built reference systems, then proposed PaperVoyager, a generation framework that explicitly models mechanisms and interaction logic. Results show the approach significantly outperforms baseline generation, opening a new way to explore how scientific findings actually work.
Published as I-WebGenBench : Evaluating Interactivity in LLM-Generated Scientific Web Applications arXiv:2606.00750
Read the original paper →