Disparidades lingüísticas en el rendimiento de un modelo de lenguaje grande en preguntas dentales en siete idiomas

Maikro Osvaldo Chávez Moya, Maikroilv@gmail.com, Leodanis Hernández Cabrera, Leodanishernandezcabrera1@gmail.com, Neisy Pérez Ramos and Neisyperezramos@gmail.com

SAP Artificial Intelligence in Dentistry, 2026

Abstract: Introduction: Large language models answer dental questions, but their accuracy across languages has not been measured. Methods: Sixty dental multiple-choice questions from MedMCQA were translated into Swahili, Hindi, Arabic, Chinese, Portuguese and Spanish with a fixed protocol and answered by the DeepSeek model at temperature zero, with English as the reference. Results: Accuracy was 61.7% in English and 61.7% in Portuguese, and it fell to 60.0% in Chinese, 58.3% in Arabic and Spanish, 55.0% in Hindi, and 26.7% in Swahili. The English-to-language gap reached 6.7 percentage points for Hindi and 35.0 points for Swahili. Conclusions: Accuracy held near the English reference in the high-resource European and Asian languages but collapsed in Swahili, so the language disparity in dental question answering is concentrated in the low-resource language rather than spread evenly across languages.

Date: 2026
References: Add references at CitEc
Citations:

Downloads: (external link)
https://southam.pub/journals/files/aid/aid2026158en.pdf (application/pdf)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:cwf:aidart:aid2026158

DOI: 10.62486/aid2026158

Access Statistics for this article

More articles in SAP Artificial Intelligence in Dentistry from South American Publishing
Bibliographic data for series maintained by South American Publishing Journals Manager ().