Reproduce “Large language model-assisted discovery of cohorts from scientific literature” 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
Moritz Sturm, Lisa M. Berg, Inken Berg, Harishny Sarma, Jasmin Hartmann, Denissa Girschik, Gemma Roig, Christine M. Freitag and 1 more
Researchers developed a configurable pipeline utilizing large language models to systematically identify named population cohorts directly from scientific literature based on specific research questions. Applied to youth aggression genetics, the system identified 44 eligible cohorts representing approximately 890,000 participants, while achieving a precision of 1.000 in LLM extraction against a human gold standard, demonstrating complementary coverage to existing cohort catalogues.
Key insights
- LLMs can effectively and reproducibly extract named cohorts from scientific literature, showing performance within the range of human inter-annotator agreement.
- A literature-based discovery framework complements existing structured cohort catalogues by identifying cohorts not readily found in curated resources.
- The configurable workflow enhances the transparency, auditability, and reproducibility of cohort identification for multi-study analyses.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.