Book chapter · 2026 · Open access
Benchmarking Small Language Models for Medical Error Identification and Correction in Clinical Notes
Ambient intelligence and smart environments · Published 2 June 2026 · DOI 10.3233/aise260032
In brief
“Benchmarking Small Language Models for Medical Error Identification and Correction in Clinical Notes” is a 2026 book chapter by Reshma Shenoy and 2 co-authors, published in Ambient intelligence and smart environments. OpenAlex counts 0 citations as of 19 November 2026. It is open access under a CC BY-NC licence: the full text is free to read at Ambient intelligence and smart environments.
- Citations
- 0
- Authors
- 3
Abstract
In clinical environments, small language models (SLM;≤7B parameters) have emerged as practical alternative to larger models due to their suitability for on-premise deployment on hospital infrastructure, patient-side devices, reduced privacy risks associated with sensitive data, and better alignment with healthcare system constraints. Despite this promise, SLM evaluation has largely focused on standard benchmark tasks such as exam-style question answering, report summarization, information extraction, and text classification that are often closely aligned with model training objectives. In this work, we benchmarked four general-purpose and three medical-domain SLM on a clinically grounded and comparatively under-explored task using the MEDEC dataset. This task requires models to detect the presence of medical errors, identify the erroneous sentence, and generate a corrected version. Across the evaluation settings considered, general-purpose SLM tend to achieve stronger overall performance than medical-domain models. However, no single model consistently performed best across all MEDEC subtasks, suggesting potential trade-offs between error detection and correction capabilities. Additionally, two medical-domain SLM showed difficulty in reliably following task instructions, resulting in very limited performance in this setting and highlighting potential robustness challenges. Finally, we observed that improvements in prompt design can meaningfully influence model performance; for example, refined prompting for Phi-3 improved its ability to detect errors, identify erroneous sentences, and generate corrections. These findings indicate that task-specific prompting and evaluation are important considerations when assessing the suitability of SLM for clinically oriented applications.(Code available at: https://github.com/ReshmaShenoy4/benchmark-slm-on-medec.git)
Cite this paper
Formatted from the catalogue record. Check names, capitals and page numbers against the paper before you submit.
APA 7
Shenoy, R., Khalilian, S., & Verhoeven, R. (2026). Benchmarking Small Language Models for Medical Error Identification and Correction in Clinical Notes. Ambient intelligence and smart environments. https://doi.org/10.3233/aise260032
MLA 9
Shenoy, Reshma, et al. “Benchmarking Small Language Models for Medical Error Identification and Correction in Clinical Notes.” Ambient intelligence and smart environments, 2026. https://doi.org/10.3233/aise260032.
Chicago
Shenoy, Reshma, Saeed Khalilian, and Richard Verhoeven. 2026. “Benchmarking Small Language Models for Medical Error Identification and Correction in Clinical Notes.” Ambient intelligence and smart environments. https://doi.org/10.3233/aise260032.
Harvard
Shenoy, R., Khalilian, S. and Verhoeven, R. (2026) ‘Benchmarking Small Language Models for Medical Error Identification and Correction in Clinical Notes’, Ambient intelligence and smart environments. doi:10.3233/aise260032.
IEEE
R. Shenoy, S. Khalilian, and R. Verhoeven, “Benchmarking Small Language Models for Medical Error Identification and Correction in Clinical Notes,” Ambient intelligence and smart environments, 2026, doi: 10.3233/aise260032.
BibTeX
@incollection{shenoy2026benchmarking,
title = {{Benchmarking Small Language Models for Medical Error Identification and Correction in Clinical Notes}},
author = {Shenoy, Reshma and Khalilian, Saeed and Verhoeven, Richard},
booktitle = {Ambient intelligence and smart environments},
publisher = {IOS Press},
year = {2026},
doi = {10.3233/aise260032},
url = {https://doi.org/10.3233/aise260032}
}
Download for a reference manager: BibTeX (.bib) · RIS for Zotero, Mendeley and EndNote (.ris)
Topic and keywords
Health Sciences › Medicine › Health Informatics › Artificial Intelligence in Healthcare and Education
- small language models
- clinical notes
- prompt engineering
- instruction following
- model robustness
- benchmarking
Work with this paper
Most-cited papers in Medicine
- Yaşlı Kadınlarda Üreme SağlığıDemet Kokanalı, Yaprak Engin Üstün · Jinekoloji-Obstetrik ve Neonatoloji Tıp Dergisi · 2019 · 169,630 citations
- Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 CountriesHyuna Sung, Adalberto M. Filho, Mathieu Laversanne and 5 more · CA A Cancer Journal for Clinicians · 2021 · 119,885 citations
- “Mini-mental state”Marshal F. Folstein, Susan E. Folstein, Paul R. McHugh · Journal of Psychiatric Research · 1975 · 90,424 citations
- Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countriesFreddie Bray, Jacques Ferlay, Isabelle Soerjomataram and 3 more · CA A Cancer Journal for Clinicians · 2018 · 87,396 citations
- Hallmarks of Cancer: The Next GenerationDouglas Hanahan, Robert A. Weinberg · Cell · 2011 · 69,758 citations
Questions about this paper
How do I cite this paper in APA style?
Shenoy, R., Khalilian, S., & Verhoeven, R. (2026). Benchmarking Small Language Models for Medical Error Identification and Correction in Clinical Notes. Ambient intelligence and smart environments. https://doi.org/10.3233/aise260032
How many times has this paper been cited?
0 times, according to OpenAlex on 19 November 2026. Google Scholar and Scopus count different sets of sources, so their numbers differ.
Is this paper free to read?
Yes. It is open access at Ambient intelligence and smart environments: sign in and open the full text from this page.
Who wrote this paper?
Reshma Shenoy, Saeed Khalilian, Richard Verhoeven.
What is the DOI of this paper?
10.3233/aise260032, permanently at https://doi.org/10.3233/aise260032. Sign in and open it from this page.
Bibliographic data from OpenAlex (CC0), last checked 19 November 2026. ResearchGuru does not host this paper; once you are signed in, the links open it at the publisher or the repository that does.