EMBL Launches AI Librarian: A Knowledge Layer for Life-Science Agents

EMBL-Rome · hf · 2026-08-03

EMBL introduced EMBL AI Librarian, a knowledge retrieval layer designed specifically for life-science AI agents. Traditional databases like Europe PMC require complex syntax and return full papers, whereas this tool lets agents ask questions in natural language and receive targeted evidence.

Orchestrated by a single LLM, the system plans subqueries, executes them via the Europe PMC search engine, and locates relevant evidence within papers. Evaluations show a >16 point improvement in Citation F1 over strong baselines on ScholarQABench. On the LitQA2 benchmark, an agent grounded in Librarian scores about 8 points higher than with web search. Code is open-sourced.

Original post →

More from Research

Research channel →