Ryuhei Miyazato

I am a 2nd year master’s student at the University of Electro-Communications, a member of Harada Laboratory.

Profile

Affiliation: The University of Electro-Communications
Position: 2nd year master’s student
Field of Research: Natural Language Processing, Machine Learning, Recommender Systems
Research Interests: I am interested in exploring the internal representations within LLMs, with a particular focus on understanding the dynamics of how beliefs and perspectives evolve and integrate over time in narrative and conversational contexts.

Period	Organization	Role/Position	Description
2025.04–Present	Ichikara Inc., Tokyo, Japan	outsourcing	Synthesising of the DPO dataset.
2024.08–2024.11	DOCOMO Innovations, Inc., California, United States	Research intern in the AI & Data Analytics Group	Personalized Book Summarization and Implementation of RAG for multiple domains documents within the company.

Period	Institution	Degree	Major/Thesis
2024.04–2026.03(expected)	The University of Electro-Communications	MSc	Design Thinking and Data Science
2020.04–2024.03	The University of Electro-Communications	BSc	Design Thinking and Data Science
2017.07-2020.03	Simizu Higashi High School	High school	Soccer Recommendation

BookAsSumQA: Evaluation of Aspect-Based Book Summarization via Question Answering, Ryuhei Miyazato, Ting-Ruen Wei, Xuyang Wu, Hsin-Tai Wu, Kei Harada, preprint, 2025
Ensembling Multiple Hallucination Detectors Trained on VLLM Internal Representations, Yuto Nakamizo, Ryuhei Miyazato(co-first-author), Hikaru Tanabe, Ryuta Yamakura, Kiori Hatanaka, KDDCup Workshop at the 31st SIGKDD Conference on Knowledge Discovery and Data Mining (KDD2025), Toronto, Canada, 2025
Book Recommendation System in University Libraries Considering Course Subjects and Users’ Learning Situation, Ryuhei Miyazato, Tetsuro Nishino, Kei Harada, Transactions on Mathematical Modeling and its Applications (TOM), 2025. https://ipsj.ixsq.nii.ac.jp/records/2000549

非直線的な物語における言語モデルの時間的推論能力の検証, 宮里龍平, 丹羽彩奈, 乾健太郎, 原田慧, 第20回言語処理若手シンポジウム(YANS2025), Sizuoka, Japan (2025.9)
大規模言語モデルを用いた関係性抽出に基づく小説のアスペクトベース要約手法の提案, 宮里龍平, Hsin-Tai Wu, 原田慧, 岡本一志, 柴田淳司, 軽部幸起, 2025年度人工知能学会全国大会 (JSAI2025), Osaka, Japan (2025.5)
⼤学図書館利⽤者の学習状況を考慮した授業関連の図書推薦⼿法の提案, 宮里龍平, 西野哲朗, 原田慧, 第148回MPS・第78回BIO合同研究発表会, Okinawa, Japan (2024.6)