Prompting language influences diagnostic reasoning and accuracy of large language models

Adrien Bazoge, Josselin Corvellec, Sofiane Djillali Sid-Ahmed, Pierre-Antoine Gourraud

arXiv:2605.19173 · 2026-05-20 공개 · arXiv · PDF

llm-evaluation clinical-decision-support multilingual-llm llm-accuracy language-effect clinical-vignettes bio-mistral prompting-language

Abstract

Large language models (LLMs) are increasingly explored for clinical decision support, yet most evaluations are conducted in English, leaving their reliability in other languages uncertain. Here we evaluate the impact of prompting language on diagnostic reasoning and final diagnosis accuracy by comparing English and French performance across five LLMs (o3, DeepSeek-R1, GPT-4-Turbo, Llama-3.1-405B-Instruct, and BioMistral-7B). A total of 180 clinical vignettes covering 16 medical specialties were assessed by two physicians using an 18-point scale evaluating both diagnosis accuracy and reasoning quality. Four of the five models performed better in English (mean difference 0.37-0.91, adjusted p < 0.05), with the gap spanning multiple aspects of reasoning, including differential diagnosis, logical structure, and internal validity. o3 was the only model showing no overall language effect. These findings demonstrate that prompting language remains a critical determinant of LLM clinical performance, with implications for equitable linguistico-cultural deployment worldwide.

한국어 요약

📋 한 줄 요약

**[임상 LLM / 다국어 평가]** 5개 LLM의 영어·프랑스어 임상 진단 추론을 비교해 프롬프트 언어가 진단 정확도와 추론 품질에 일관된 영향을 미침을 보고.

🎯 핵심 기여도

💡 핵심 아이디어

"같은 모델, 같은 환자, 다른 언어"가 다른 임상 판단을 낳는다. 언어 자체가 LLM 임상 신뢰성의 결정 변수임을 인정해야 한다.

🔬 기술적 접근법

📊 주요 결과

💭 의의 및 한계

**의의**: 글로벌 임상 배포에서 언어가 안전·형평성의 핵심 축임을 정량 데이터로 입증, 다언어·다문화 평가의 의무화 근거 제공. **한계**: 영어·프랑스어 두 언어와 16개 전문분야로 한정, 더 다양한 언어·문화·임상 환경으로의 확장은 후속 과제.

🚀 실용적 활용