Can Continual Pre-training Bridge the Performance Gap between General-purpose and Specialized Language Models in the Medical Domain?

TL;DR

This study demonstrates that continual pre-training and merging of language models with domain-specific data significantly improve their performance in German medical tasks, making smaller models more competitive.

Contribution

It introduces a new methodology for domain adaptation of LLMs in German medical language, including creating a high-quality corpus and merging models to enhance specialization.

Findings

Specialized 7B models outperform general-purpose models on German medical benchmarks.

Domain adaptation increases win-rate of smaller models against larger ones by approximately 3.5 times.

Model merging restores instruction-following but introduces language mixing and verbosity.

Abstract

This paper narrows the performance gap between small, specialized models and significantly larger general-purpose models through domain adaptation via continual pre-training and merging. We address the scarcity of specialized non-English data by constructing a high-quality German medical corpus (FineMed-de) from FineWeb2. This corpus is used to continually pre-train and merge three well-known LLMs (ranging from $7B$ to $24B$ parameters), creating the DeFineMed model family. A comprehensive evaluation confirms that specialization dramatically enhances $7B$ model performance on German medical benchmarks. Furthermore, the pairwise win-rate analysis of the Qwen2.5-based models demonstrates an approximately $3.5$-fold increase in the win-rate against the much larger Mistral-Small-24B-Instruct through domain adaptation. This evidence positions specialized $7B$ models as a competitive, resource-efficient solution for complex medical instruction-following tasks. While model merging successfully restores instruction-following abilities, a subsequent failure mode analysis reveals inherent trade-offs, including the introduction of language mixing and increased verbosity, highlighting the need for more targeted fine-tuning in future work. This research provides a robust, compliant methodology for developing specialized LLMs, serving as the foundation for practical use in German-speaking healthcare contexts.