Exportar Publicação

Marcelo West Moreira, José Amancio Macedo Santos, Brito e Abreu, F., Manoel Mendonça & Glauco de Figueiredo Carneiro (2026). Benchmarking Small Language Models for Medical Question Answering in Resource-Constrained Environments. In Proceedings of the 28th International Conference on Enterprise Information Systems. (pp. 477-488). Benidorm, Spain: SCITEPRESS - Science and Technology Publications.

M. West et al.,  "Benchmarking Small Language Models for Medical Question Answering in Resource-Constrained Environments", in Proc. of the 28th Int. Conf. on Enterprise Information Systems, Benidorm, Spain, SCITEPRESS - Science and Technology Publications, 2026, pp. 477-488

@inproceedings{west2026_1784752298861,
	author = "Marcelo West Moreira and José Amancio Macedo Santos and Brito e Abreu, F. and Manoel Mendonça and Glauco de Figueiredo Carneiro",
	title = "Benchmarking Small Language Models for Medical Question Answering in Resource-Constrained Environments",
	booktitle = "Proceedings of the 28th International Conference on Enterprise Information Systems",
	year = "2026",
	editor = "",
	volume = "",
	number = "",
	series = "",
	doi = "10.5220/0014737700004018",
	pages = "477-488",
	publisher = "SCITEPRESS - Science and Technology Publications",
	address = "Benidorm, Spain",
	organization = "INSTICC",
	url = "https://iceis.scitevents.org/?y=2026"
}

TY  - CPAPER
TI  - Benchmarking Small Language Models for Medical Question Answering in Resource-Constrained Environments
T2  - Proceedings of the 28th International Conference on Enterprise Information Systems
AU  - Marcelo West Moreira
AU  - José Amancio Macedo Santos
AU  - Brito e Abreu, F.
AU  - Manoel Mendonça
AU  - Glauco de Figueiredo Carneiro
PY  - 2026
SP  - 477-488
DO  - 10.5220/0014737700004018
CY  - Benidorm, Spain
UR  - https://iceis.scitevents.org/?y=2026
AB  - While Large Language Models (LLMs) demonstrate significant potential in identifying complex patterns, their deployment in the medical domain is restricted by critical barriers: the misalignment between standardized benchmarks and real-world patient consultations, the data privacy risks inherent in commercial APIs, and the prohibitive computational costs of local deployment. This study investigates the feasibility of Small Language Models (SLMs)—defined here as architectures with fewer than 10 billion parameters—as a viable solution to these constraints. We conduct a comparative evaluation of domain-specific and general-purpose models from the Hugging Face ecosystem, analyzing their effectiveness in medical question-answering tasks specifically tailored for privacy-conscious and resource-limited settings. Our results reveal a distinct trade-off between information density and factual reliability; while general-purpose models achieved higher comprehensiveness at the cost of increased hallucinations, clinically conservative models minimized errors but lacked coverage. Ultimately, we identify that balanced architectures, such as Medgemma-1.5-4b, offer the most promising path for deploying safe, accessible, and efficient AI decision support in resource-constrained medical environments.
ER  -