Reproduce “Story2Proposal: A Scaffold for Structured Scientific Paper Writing” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Zhuoyang Qian, Wei Shi, Xu Lin, Li Ling, Meng Luo, Ziming Wang, Zhiwei Zhang, Tengyue Xu and 14 more

The AgentAlpha Team developed Story2Proposal, a contract-governed multi-agent framework that structures scientific paper writing to ensure document-wide integrity and alignment between text and visuals. This system achieved an average expert evaluation score of 6.145, outperforming an unconstrained baseline by +2.182 points and a structured baseline by +0.508 points.

Key insights

  • A contract-governed multi-agent architecture significantly enhances the reliability and structural integrity of automatically generated scientific manuscripts.
  • Explicitly modeling and enforcing structural and visual obligations via a shared visual contract prevents common failure modes like structural drift and missing visual artifacts.
  • The 'generate–evaluate–adapt' loop allows for continuous validation and correction, ensuring consistency and alignment between narrative reasoning, experimental evidence, and visual artifacts throughout the writing process.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.