Journal Watch · Nature · 2026

Synthesizing scientific literature with retrieval-augmented language models.

We open-source all artefacts, including our code, models, data store, datasets and a public demo.

Asai A, He J, Shao R et al.Nature3 min readPharma relevance 40/100

DOI: 10.1038/s41586-025-10072-4PMID: 41639446Study type: research study

TL;DR

We open-source all artefacts, including our code, models, data store, datasets and a public demo.

Key findings

  • OpenScholar's data store, retriever and self-feedback inference loop improve off-the-shelf LMs: for instance, OpenScholar-GPT-4o improves the correctness of GPT-4o by 12%.
  • In human evaluations, experts preferred OpenScholar-8B and OpenScholar-GPT-4o responses over expert-written ones 51% and 70% of the time, respectively, compared with 32% for GPT-4o.
  • We open-source all artefacts, including our code, models, data store, datasets and a public demo.

Numbers at a glance (from the abstract)

6.1%5.5%90%12%51%70%32%

Every figure above is taken verbatim from the abstract below. Check the full text for denominators, confidence intervals and subgroup detail.

What they did

To evaluate OpenScholar, we develop ScholarQABench, the first large-scale multi-domain benchmark for literature search, comprising 2,967 expert-written queries and 208 long-form answers across computer science, physics, neuroscience and biomedicine. Despite being a smaller open model, OpenScholar-8B outperforms GPT-4o by 6.1% and PaperQA2 by 5.5% in correctness on a challenging multi-paper synthesis task from the new ScholarQABench. Although GPT-4o hallucinates citations 78-90% of the time, OpenScholar achieves citation accuracy on par with human experts.

Full abstract (original)

Background: Scientific progress depends on the ability of researchers to synthesize the growing body of literature. Can large language models (LLMs) assist scientists in this task? Here we introduce OpenScholar, a specialized retrieval-augmented language model (LM)<sup>1</sup> that answers scientific queries by identifying relevant passages from 45 million open-access papers and synthesizing citation-backed responses.

What they did: To evaluate OpenScholar, we develop ScholarQABench, the first large-scale multi-domain benchmark for literature search, comprising 2,967 expert-written queries and 208 long-form answers across computer science, physics, neuroscience and biomedicine. Despite being a smaller open model, OpenScholar-8B outperforms GPT-4o by 6.1% and PaperQA2 by 5.5% in correctness on a challenging multi-paper synthesis task from the new ScholarQABench. Although GPT-4o hallucinates citations 78-90% of the time, OpenScholar achieves citation accuracy on par with human experts.

Results & conclusions: OpenScholar's data store, retriever and self-feedback inference loop improve off-the-shelf LMs: for instance, OpenScholar-GPT-4o improves the correctness of GPT-4o by 12%. In human evaluations, experts preferred OpenScholar-8B and OpenScholar-GPT-4o responses over expert-written ones 51% and 70% of the time, respectively, compared with 32% for GPT-4o. We open-source all artefacts, including our code, models, data store, datasets and a public demo.

Abstract © the authors / Nature (2026). Reproduced here for discussion and indexing, with attribution. Read the original at https://doi.org/10.1038/s41586-025-10072-4 · PubMed 41639446.

Why it matters for pharma & biotech

  • Relevant to General & Medicine teams tracking Nature for new targets, mechanisms and biomarkers.
  • As earlier-stage work, its near-term value is target validation and assay design rather than clinical decisions.
  • For searchers: the full abstract is reproduced below, so you can judge relevance before opening the publisher page.

Limitations & what to check in the full text

  • This brief is based on the published abstract only. Effect sizes, subgroup results and safety detail in the full text may qualify the headline findings.

Frequently asked questions

What did this Nature study find?

We open-source all artefacts, including our code, models, data store, datasets and a public demo.

What type of study is this?

It is a research study, published in Nature in 2026. To evaluate OpenScholar, we develop ScholarQABench, the first large-scale multi-domain benchmark for literature search, comprising 2,967 expert-written queries and 208 long-form answers across computer science, physics, neuroscience and biomedicine.

Where can I read the original paper?

The original is at the publisher via DOI 10.1038/s41586-025-10072-4. PubMed ID: 41639446. BioStudioAI reproduces the abstract for discussion and links to the source.

How to cite the original: Asai A, He J, Shao R et al.. Synthesizing scientific literature with retrieval-augmented language models.. Nature (2026). DOI: 10.1038/s41586-025-10072-4.
This brief: BioStudioAI Journal Watch, 29 October 2026. https://biostudioai.com/blog/papers/synthesizing-scientific-literature-with-retrieval-augmented-language-mod-100724/

Reading papers is step one. Analysing the data is the work.

BioStudioAI turns raw lab files into QC'd results, statistics, figures, reports and slides — hosted on your servers, every number computed, nothing typed.

Open BioStudioAI More guides