Overview
PaperOrchestra is a multi-agent system from Google researchers that turns unstructured research materials into a LaTeX manuscript ready for submission, including a literature review and generated figures. The researcher supplies an idea summary, an experimental log, the conference’s LaTeX template and its guidelines. Five agents then outline the paper, make the plots, review the literature, write the sections and refine the manuscript using simulated peer-review feedback.
To evaluate it, the authors built a benchmark from the raw materials of 200 papers from leading AI conferences, reverse-engineered from the published papers. The paper was posted on arXiv on 6 April 2026 and is published in the proceedings of the Third Conference on Language Modeling (2026).
Benefits and outcomes
- The authors report that in side-by-side human evaluations PaperOrchestra beat autonomous baselines by absolute win-rate margins of 50% to 68% for literature review quality and 14% to 38% for overall manuscript quality.
Notes and quotes
- Greg De Temmerman notes that the paper lists limitations, including hallucinations in the generated figures, a gap in writing quality and omissions of strong references in the field.
Sources
- PaperOrchestra: A Multi-Agent Framework for Automated AI Research Paper Writing, Yiwen Song, Yale Song, Tomas Pfister and Jinsung Yoon, arXiv, 6 April 2026
- PaperOrchestra, Google Research publication page
- On PaperOrchestra: AI Manuscript Generation and the Authorship Problem, Scribendi
- Science in auto-pilot?, Greg De Temmerman, LinkedIn, November 2026
This entry was written with the help of AI, from the sources listed. Please check the sources before relying on it, and let us know if anything needs correcting.
