Bridging Language Barriers in Healthcare: A Study on Arabic LLMs

TL;DR

The paper tackles the difficulty of creating Arabic medical LLMs capable of multilingual understanding and domain knowledge, showing that translating data alone does not guarantee strong Arabic clinical performance. It systematically evaluates open-source models on Arabic medical benchmarks, adapts evaluation tools for RTL Arabic, and experiments with translation pipelines and bilingual fine-tuning. Key findings reveal that optimal language mixtures are task-dependent, larger models benefit more from Arabic-focused pretraining, and fine-tuning alone often fails to surpass pretraining, underscoring the need for substantial pretraining data and careful data curation. The work provides practical guidance for building inclusive medical AI systems for Arabic-speaking communities and highlights the importance of task-aware data strategies and robust evaluation benchmarks.

Abstract

This paper investigates the challenges of developing large language models (LLMs) proficient in both multilingual understanding and medical knowledge. We demonstrate that simply translating medical data does not guarantee strong performance on clinical tasks in the target language. Our experiments reveal that the optimal language mix in training data varies significantly across different medical tasks. We find that larger models with carefully calibrated language ratios achieve superior performance on native-language clinical tasks. Furthermore, our results suggest that relying solely on fine-tuning may not be the most effective approach for incorporating new language knowledge into LLMs. Instead, data and computationally intensive pretraining methods may still be necessary to achieve optimal performance in multilingual medical settings. These findings provide valuable guidance for building effective and inclusive medical AI systems for diverse linguistic communities.