DeLTa: A Decoding Strategy based on Logit Trajectory Prediction Improves Factuality and Reasoning Ability
Second Workshop on Uncertainty-Aware NLP, EMNLP 2025
Ph.D. student at Kyoto University, Graduate School of Informatics, advised by Hidetoshi Shimodaira. I work on large language models — hallucination, decoding, post-training, and alignment.
I'm a first-year Ph.D. student in the Data Science Course at Kyoto University, supported by the DoGS Fellowship (2025–2028). Before that I completed my Master's in the same group, and my Bachelor's at Chuo University in Tokyo.
I've also spent time as an LLM R&D intern at Spiral AI and on Rakuten's AI Architecture team. My research has been recognized with a Sponsor Award at ANLP 2024.
Second Workshop on Uncertainty-Aware NLP, EMNLP 2025
23rd Workshop on Biomedical NLP, ACL 2024
Association for Natural Language Processing (ANLP) 2024Sponsor Award
Ph.D., Informatics (Data Science)
Kyoto University · Graduate School of Informatics
Advisor: Hidetoshi Shimodaira. Expected completion 2028. Supported by the DoGS Fellowship.
M.S., Informatics (Data Science)
Kyoto University · Graduate School of Informatics
Advisor: Hidetoshi Shimodaira.
B.E., Information and System Engineering
Chuo University · Faculty of Science and Engineering
Research Assistant
Graduate School of Informatics, Kyoto University
Part-time, AI Architecture
Rakuten Group · AI & Data Division
AI for Business Department, AI Architecture Section.
Intern, LLM Deep Learning Engineer
Spiral AI · R&D
DoGS Fellow
Division of Graduate Studies, Kyoto University
Research stipend ¥250,000/month + ¥900,000/year research grant.
Sponsor Award
Association for Natural Language Processing (ANLP) 2024