Reproduce “Agentic AI-Empowered Dynamic Survey Framework” 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
Furkan Mumcu, Lokman Bekit, Michael J. Jones, Anoop Cherian, Yasin Yilmaz
An agentic AI framework was developed to enable continuous updating and maintenance of scientific survey papers, reframing survey writing as a long-horizon maintenance problem. The system integrates new research incrementally while preserving original structure and style, achieving high update quality and conservativeness compared to unconstrained LLM baselines. Human evaluations indicate the generated updates are comparable to human-written edits.
Key insights
- Reframing survey writing from a one-time generation task to a 'long-horizon maintenance problem' allows for continuous, adaptive knowledge synthesis.
- Decomposing the complex task of survey updating into specialized, constrained agent roles leads to significantly better control and quality compared to monolithic LLM approaches.
- Preserving an immutable survey structure and performing highly localized, conservative edits are crucial for maintaining document integrity, coherence, and stylistic consistency over time.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.