Our study, “Fine-Tuning LLMs for Legal Knowledge Assessment: A Taiwan Case Study,” was invited by the Legal Design Lab at Stanford Law School to be presented at the AI & Access to Justice Webinar on June 6, 2025. Following the presentation, members of the Lab helped adapt and rewrite the work into an article published on Medium, introducing Taiwan’s experience to the global academic and legal communities.

This research examines how open-source language models can be fine-tuned with locally sourced legal data in the Taiwan context, so that they can better understand and respond to professional legal knowledge—particularly exam-style questions requiring legal reasoning and judgment. Our findings demonstrate that the integration of local legal corpora significantly improves model performance, underscoring the importance of domain- and jurisdiction-specific data for any LLM seeking to operate in the legal domain.

The Taiwan case therefore offers a valuable pathway for the integration of AI and legal education, with broader implications for access to justice and legal innovation.

👉 Read the full Medium article:
https://medium.com/legal-design-and-innovation/fine-tuning-llms-for-legal-knowledge-assessment-a-taiwan-case-study-7cee320fd3d1