Author: Toshiki Demizu (出水 利樹) — GitHub/Hugging Face ID: @demimomi
Affiliation: ソフトバンク株式会社、MONET Technologies株式会社
Course: Large Language Model Development Lecture (Autumn 2025)
Special thanks to the instructors and collaborators at the Matsuo Laboratory and SoftBank Corp. for their guidance in practical LLM research and fine-tuning experiments.
Project Background
This model was developed as part of the Large Language Model Development Lecture (大規模言語モデル開発講座 基礎編) at the Matsuo Laboratory, The University of Tokyo.
基礎編4,000名参加 修了率46.9% (メモ:応用編は3,800名が参加)
The course provides practical training in fine-tuning, preference optimization (DPO), and deployment of large-scale Japanese language models.
More details about the course can be found here:
🔗 The University of Tokyo WebLab – Large Language Model Course
This specific fine-tuning ("sarashina/dpo") was carried out using the base model sbintuitions/sarashina2.2-0.5b
as part of the course’s applied experiments.