EMBL AI Librarian: Life-Sciences Knowledge Layer for AI Agents

Source

Source name: arXiv
Source kind: paper
Authors / organization: Luigi Sigillo, Matteo Silvestri, Francesco Tabaro, Rajat Bhatnagar, Syed Irtaza Mubashar, Matt Jeffryes, Daljit Nijjer, Vittorio Perera, Ola Spjuth, Julio Saez-Rodriguez, Melissa Harrison, Fabio Petroni
Venue: arXiv: cs.CL
Published: 2026-07-30
Collected: 2026-07-31
URL: http://arxiv.org/abs/2607.28229v1
PDF: https://arxiv.org/pdf/2607.28229v1
DOI: UNKNOWN

Plain-Language Brief

The paper introduces EMBL AI Librarian, a tool that allows AI agents to access and process life sciences literature more effectively by asking natural language questions and receiving relevant evidence in response. The tool is evaluated across several benchmarks and improves performance on various tasks compared to existing methods.

What the Document Says

The web is increasingly accessed by AI agents, particularly in the life-sciences field where agentic pipelines are growing.
Accessing literature is crucial for these AI agents, and resources like Europe PMC with over 40M indexed records are commonly used.
However, these resources were not built for AI agents, requiring them to learn complex syntax and read full papers to find evidence.
The paper introduces EMBL AI Librarian, which upgrades the Europe PMC interface for AI agents, allowing them to ask natural language questions and receive evidence that answers them.
A single LLM (Langauge Model) orchestrates the knowledge retrieval process by planning complementary subqueries executed by the live Europe PMC search engine, reading selected papers, and locating relevant evidence.
The paper evaluates Librarian across four benchmarks: literature synthesis, claim verification, open-domain question answering, and downstream biology tasks such as protocol questions and sequence manipulation.
Librarian improves Citation F1 by more than $16$ points over strong recently published baselines on ScholarQABench.
When used as the retrieval layer of an existing claim-verification pipeline, it increases agreement with expert consensus.
On the open-form LitQA2 benchmark, a GPT-5.4 agent scores about $8$ points higher when grounded in Librarian than with web search.
The paper concludes that equipping life-science agents with the Librarian knowledge layer improves performance across a range of tasks.

Key Concepts Mentioned

AI agents
life-sciences literature
Europe PMC
EMBL AI Librarian
natural language processing
evaluation benchmarks
Citation F1
GPT-5.4

Claims Made by the Document

EMBL AI Librarian improves performance across a range of tasks compared to existing methods.
Using EMBL AI Librarian as the retrieval layer of an existing claim-verification pipeline increases agreement with expert consensus.

Evidence Used

Improvement in Citation F1 by more than $16$ points over strong recently published baselines.
A GPT-5.4 agent scoring about $8$ points higher when grounded in Librarian than with web search.

Methods / Approach

Using a single LLM to orchestrate the knowledge retrieval process

Limitations Stated by the Authors

Not stated in source

Limitations Noted by the Agent

Agent interpretation — not from the source.

Unknown, as the paper focuses on evaluating EMBL AI Librarian rather than its limitations.

Open Questions

What specific improvements does EMBL AI Librarian bring to each task it is evaluated on?
How does EMBL AI Librarian handle ambiguity in natural language queries?

Notes for Future Q&A

May include interpretation. Labeled where it does.

The paper provides evaluation results for EMBL AI Librarian but does not explicitly discuss its limitations or how it handles ambiguity.

Raw Abstract / Excerpt

The web is increasingly accessed by AI agents rather than humans. Every agent needs knowledge, especially in the life-sciences, where agentic pipelines are growing fast. Access to the literature is a crucial part of that need, and resources such as Europe PMC, with over 40M indexed records, are widely used to meet it. Yet these resources were not built for AI agents: they take keywords and complex syntax and return whole papers, so every agent must learn the syntax, issue several searches, and read full papers to find the evidence it needs. We introduce EMBL AI Librarian, a knowledge layer that upgrades the Europe PMC interface for AI agents: an agent asks in natural language and receives evidence that answers it. A single LLM orchestrates the whole knowledge retrieval process: it plans complementary subqueries executed by the live Europe PMC search engine, then reads the selected papers and locates the relevant evidence. We evaluate Librarian across four benchmarks: literature synthesis, claim verification, open-domain question answering, and downstream biology tasks such as protocol questions and sequence manipulation. On ScholarQABench, Librarian improves Citation F1 by more than $16$ points over strong recently published baselines. Used as the retrieval layer of an existing claim-verification pipeline, it increases agreement with expert consensus; and on the open-form LitQA2 benchmark, a GPT-5.4 agent scores about $8$ points higher when grounded in Librarian than with web search. Overall, our results show that equipping life-science agents with the Librarian knowledge layer improves performance across a range of tasks. We release our code publicly at https://github.com/petroni-lab/librarian