GLiNER-BioMed: a suite of efficient models for open biomedical named entity recognition

Abstract

MOTIVATION: Biomedical named entity recognition (NER) presents unique challenges due to specialized vocabularies, the sheer volume of entities, and the continuous emergence of novel entities. Traditional NER models, constrained by fixed taxonomies and human annotations, struggle to generalize beyond predefined entity types. RESULTS: To address these issues, we introduce GLiNER-BioMed, a domain-adapted suite of GLiNER models for biomedicine. Our approach first distills the annotation capabilities of large language models (LLMs) into a smaller, more efficient model, enabling the generation of high-coverage biomedical NER data. We subsequently train two GLiNER architectures, uni- and bi-encoder, at multiple scales to balance computational efficiency and performance. Experiments on eight biomedical datasets demonstrate that GLiNER-BioMed achieved state-of-the-art zero-shot performance (micro-F1 59.77%), exceeding the strongest baseline by 5.96 points (P < .001). In few-shot learning, the bi-encoder variant reached 70.39% (10-shot), consistently outperforming the strongest baseline across all settings (P < .05). Our findings show that the uni-encoder GLiNER-BioMed achieves the strongest zero-shot performance, while the bi-encoder offers superior few-shot gains and substantially higher inference throughput (+39%-568%), making it well-suited to annotation-limited, latency-sensitive, or large-label-space settings. Ablation studies further indicate that combining synthetic biomedical pre-training with general-domain post-training is essential for capturing domain-specific knowledge while maintaining precision-recall balance. AVAILABILITY AND IMPLEMENTATION: The source code, datasets, and models are publicly available at https://github.com/ds4dh/GLiNER-biomed.

DICD-IE: an expert-annotated benchmark for evaluating pretrained biomedical language models in dental implant literature and clinical knowledge extraction
Almonzer Salah Nooraldaim, Azza A. Abushama, Hiba Ahmed +7
Scientific Reports30 Sep 2026Peer-reviewedCC BY-NC-ND 4.0Full textMolecular biology & gene regulationCrossref
Autonomous biomedical research with an artificial intelligence agent
Kexin Huang, Serena Zhang, Hanchen Wang +25
Science9 Jul 2026Peer-reviewedLicence not statedFull text195 citedMolecular biology & gene regulationOpenAlex
Bridging communication gaps across the biomedical ecosystem
Elizabeth A. Burke, Samantha N. Schumm, Amy Jobe +2
Frontiers in Systems Biology12 Oct 2026Peer-reviewedCC BY 4.0Full textMolecular biology & gene regulationPMCEurope PMC
The transformative impact of stem cell core facilities in biomedical research
Barbara Corneo, Giovanni Faga, Inês Figueira +12
Stem Cell Reports5 Oct 2026Peer-reviewedCC BY-NC-ND 4.0Full textMolecular biology & gene regulationHAL
Efficacy of statistical and design interventions for improving reproducibility in biomedical research
Bernhard Voelkl, Rachel Heyard, Laura Caquelin +3
BMC Biology21 Oct 2026Peer-reviewedCC BY 4.0Molecular biology & gene regulation General biologyCrossref

Cite

BibTeX RIS

BibTeX

@article{yazdani2026glinerbiomed,
  title = {{GLiNER-BioMed: a suite of efficient models for open biomedical named entity recognition}},
  author = {Anthony Yazdani and Ihor Stepanov and Douglas Teodoro},
  journal = {Bioinformatics},
  year = {2026},
  month = may,
  doi = {10.1093/bioinformatics/btag322},
  url = {https://doi.org/10.1093/bioinformatics/btag322},
  note = {CC BY},
}

RIS

TY  - JOUR
TI  - GLiNER-BioMed: a suite of efficient models for open biomedical named entity recognition
AU  - Anthony Yazdani
AU  - Ihor Stepanov
AU  - Douglas Teodoro
PY  - 2026
DA  - 2026/05/19
JO  - Bioinformatics
DO  - 10.1093/bioinformatics/btag322
UR  - https://doi.org/10.1093/bioinformatics/btag322
AB  - MOTIVATION: Biomedical named entity recognition (NER) presents unique challenges due to specialized vocabularies, the sheer volume of entities, and the continuous emergence of novel entities. Traditional NER models, constrained by fixed taxonomies and human annotations, struggle to generalize beyond predefined entity types. RESULTS: To address these issues, we introduce GLiNER-BioMed, a domain-adapted suite of GLiNER models for biomedicine. Our approach first distills the annotation capabilities of large language models (LLMs) into a smaller, more efficient model, enabling the generation of high-coverage biomedical NER data. We subsequently train two GLiNER architectures, uni- and bi-encoder, at multiple scales to balance computational efficiency and performance. Experiments on eight biomedical datasets demonstrate that GLiNER-BioMed achieved state-of-the-art zero-shot performance (micro-F1 59.77%), exceeding the strongest baseline by 5.96 points (P < .001). In few-shot learning, the bi-encoder variant reached 70.39% (10-shot), consistently outperforming the strongest baseline across all settings (P < .05). Our findings show that the uni-encoder GLiNER-BioMed achieves the strongest zero-shot performance, while the bi-encoder offers superior few-shot gains and substantially higher inference throughput (+39%-568%), making it well-suited to annotation-limited, latency-sensitive, or large-label-space settings. Ablation studies further indicate that combining synthetic biomedical pre-training with general-domain post-training is essential for capturing domain-specific knowledge while maintaining precision-recall balance. AVAILABILITY AND IMPLEMENTATION: The source code, datasets, and models are publicly available at https://github.com/ds4dh/GLiNER-biomed.
SN  - 1367-4803
SN  - 1367-4811
ER  -

Abstract

Related

Cite