I-WebGenBench : Evaluating Interactivity in LLM-Generated Scientific Web Applications

2026-05-30

AI Summary

This work proposes PaperVoyager, an end-to-end paper-to-interactive-system agent that transcends the limitations of existing document intelligence systems, which typically convert scientific papers into static content and fail to capture their dynamic mechanisms and state transitions. By integrating vision-language models, system modeling, and web synthesis techniques, PaperVoyager automatically transforms PDF-based research papers into interactive web systems that support user-driven operations and provide real-time dynamic feedback. To evaluate such capabilities, we introduce I-WebGenBench, the first benchmark tailored for interactive scientific web generation, and develop a structured generation framework that explicitly models system mechanics and interaction logic. Experimental results on 19 expert-annotated papers demonstrate that PaperVoyager significantly outperforms baseline approaches, establishing a new paradigm for interactive understanding of scientific literature.

Abstract

Recent advances in visual language models have enabled autonomous agents for complex reasoning, tool use, and document understanding. However, existing document agents mainly transform papers into static artifacts such as summaries, webpages, or slides, which are insufficient for technical papers involving dynamic mechanisms and state transitions. In this work, we propose a Paper-to-Interactive-System Agent that converts research papers into executable interactive web systems. Given a PDF paper, the agent performs end-to-end processing without human intervention, including paper understanding, system modeling, and interactive webpage synthesis, enabling users to manipulate inputs and observe dynamic behaviors. To evaluate this task, we introduce a benchmark of 19 research papers paired with expert-built interactive systems as ground truth. We further propose PaperVoyager, a structured generation framework that explicitly models mechanisms and interaction logic during synthesis. Experiments show that PaperVoyager significantly improves the quality of generated interactive systems, offering a new paradigm for interactive scientific paper understanding.

Problem

• interactive systems
• scientific papers
• dynamic mechanisms
• state transitions
• document understanding

Innovation

• interactive web generation
• paper-to-system conversion
• structured generation framework
• scientific document understanding
• autonomous agent