• Sep. 17, 2024:

📰 Featured in ITmedia

"Why Teach AI ‘Japanese’? 

Experiment and Analysis with 35 Types of LLMs: 

Research Report by Tokyo Tech and Others"

• Sep. 6-7, 2024:

🎉🏆 Won the ‘Encouragement Award’ 

and ‘Sponsor Award’

at YANS2024

• Sep. 3, 2024:

🎉🏆 Awarded Excellence

at the 261st NL Research Presentation

Refereed Publications 📜

International Conferences and Symposiums 🌏

Non-Refereed Publications 📝

Domestic Conferences and Symposiums 🇯🇵

• Sep. 6-7, 2024:

🎉🏆 Won the ‘Encouragement Award’ 

and ‘Sponsor Award’

at YANS2024

「Easily detectable LLMs without sacrificing its generative capability」[Poster (JP)]
Koshiro Saito [1], Ryuto Koike [1], Masahiro Kaneko[2], Naoaki Okazaki[1] (1: TITECH, 2: MBZUAI)

• Sep. 3, 2024:

🎉🏆 Awarded Excellence

at the 261st NL Research Presentation

「Advantages of Training LLMs on Japanese Text」[Paper (JP), Slides (JP)]
Koshiro Saito [1], Sakae Mizuki [2,1], Masanari Ohi [1], Takashi Nakamura [1], Taihei Shiotani [1], Koki Maeda [1], Ma Youmi [1], Kakeru Hattori [1], Kazuki Fujii [1], Takumi Okamoto [1], Shigeki Ishida [1], Hiroya Takamura [2], Rio Yokota [1], Naoaki Okazaki [1] (1: TITECH, 2: AIST)

• Jul. 2024 -  Present:

RA Okazaki Lab, Tokyo Tech

• Feb. 2023 - Present:

Intern Canserscan Inc.

I believe that NLP is the essential bridge that will enable the realization of a symbiotic society between AI and humans.

• NLP

  • Watermark (Currently🏃)

    • "Is it possible to make models detectable while keeping their abilities?"

    • "Is it possible to tell human-written or AI-generated, and when would it be useful?" 

• • Models/Datasets Evaluation (Currently🏃)

    • "How & What to evaluate for better models?"

    • "What are 'good' datasets?"

• • Biased LLM

    • "How to eliminate/detect biased opinions of LLMs gained by open sources?"

    • "Can models include multifaceted and diverse opinions in one response?"

• • Democratization of LLM

    • "How to develop light, highly convertible models?"

    • "How can plagiarism of models by unauthorized persons or entities be prevented?"

• • Multimodal LLM

• Apr. 2021 - Present:

Bachelor of Engineering 

(Computer Science)

Tokyo Institute of Technology

• Sep. 17, 2024:

📰 Featured in ITmedia

"Why Teach AI ‘Japanese’? 

Experiment and Analysis with 35 Types of LLMs: 

Research Report by Tokyo Tech and Others"

• Jul. 2024:

Scholarship 

by Keyence Foundation