LegalMidm: Use-Case-Driven Legal Domain Specialization for Korean Large Language Model

2026-04-28

AI Summary

This study addresses the insufficient accuracy and reliability of general-purpose large language models (LLMs) in Korean legal practice. To overcome these limitations, the authors propose a systematic, use-case-driven training framework that innovatively integrates deep involvement from legal practitioners throughout the data construction process. By leveraging high-quality Korean legal datasets, domain-adaptive fine-tuning, and expert-in-the-loop annotation and validation, they develop LegalMidm—a specialized LLM tailored for the Korean legal domain. Experimental results demonstrate that LegalMidm significantly outperforms general-purpose models across multiple critical legal tasks, thereby enhancing both practical utility and reliability in real-world legal scenarios.

Abstract

In recent years, the rapid proliferation of open-source large language models (LLMs) has spurred efforts to turn general-purpose models into domain specialists. However, many domain-specialized LLMs are developed using datasets and training protocols that are not aligned with the nuanced requirements of real-world applications. In the legal domain, where precision and reliability are essential, this lack of consideration limits practical utility. In this study, we propose a systematic training framework grounded in the practical needs of the legal domain, with a focus on Korean law. We introduce LegalMidm, a Korean legal-domain LLM, and present a methodology for constructing high-quality, use-case-driven legal datasets and optimized training pipelines. Our approach emphasizes collaboration with legal professionals and rigorous data curation to ensure relevance and factual accuracy, and demonstrates effectiveness in key legal tasks.

Problem

• legal domain
• large language models
• domain specialization
• use-case-driven
• practical utility

Innovation

• use-case-driven
• legal domain specialization
• Korean LLM
• high-quality dataset curation
• collaborative data annotation