Ryuto Koike  (小池 隆斗)
Ph.D. candidate at Tokyo Institute of Technology

EmailLinkLinkGitHubLinkedInTwitter

Hello, I am a Ph.D. candidate at Okazaki Laboratory, Department of Computer Science, Tokyo Institute of Technology. I am honored to be advised by Professor Naoaki Okazaki and Postdoc Masahiro Kaneko.
 My current research interest is centered on improving the trustworthiness of Large-scale Language Models (LLMs). I have been recently working on increasing the robustness of LLM-generated text detection.
Actively looking for research collaborations, internships, and opportunities as a visiting student. Please feel free to contact me.

✨ News 

Dec 2023: A paper on LLM detection accepted at AAAI 2024🇨🇦.

Oct 2023: Our work on LLM detection was featured in Nikkei🗞.

Aug 2023: Received Double Sponsorship Awards at the YANS 2023🇯🇵.

Apr 2023: Joined Okazaki Lab, Tokyo Institute of Technology🇯🇵.

👨‍💻 Work Experiences

Jan 2024 - Present:                 Research Engineer                                                     3keigo.com

Apr 2023 - Present:                Research Assistant                                                    Tokyo Tech

Feb 2022 - Mar 2022:            Machine Learning Engineer Intern                           Exawizards, Inc.

Sep 2021 - Jan 2022:             Research Intern                                                          CyberAgent, Inc.

Jul 2021 - Aug 2021:             Machine Learning Engineer Intern                           CyberAgent, Inc.

📜 Refereed Publications

2024

Likelihood-based Mitigation of Evaluation Bias in Large Language Models

Masanari Ohi, Masahiro Kaneko, Ryuto Koike, Mengsay Loem, Naoaki Okazaki
arXiv (under review)
[arXiv]

OUTFOX: LLM-generated Essay Detection through In-context Learning with Adversarially Generated Examples 

Ryuto Koike, Masahiro Kaneko, Naoaki Okazaki
AAAI 2024 (Acceptance rate: 18.4%)
[arXiv] [Paper] [Technical Appendix] [Video] [Data&Code]

2023

How You Prompt Matters! Even Task-Oriented Constraints in Instructions Affect LLM-Generated Text Detection

Ryuto Koike, Masahiro Kaneko, Naoaki Okazaki
arXiv (under review)
[arXiv] [Data]

2022

A Support System for Generating YouTube Titles using Style Transfer

Ryuto Koike, Masafumi Hagiwara
Transactions of Japan Society of Kansei Engineering (In Japanese)
[Paper]

🗣 Domestic Conference and Symposium (In Japanese)

Masanari Ohi, Masahiro Kaneko, Ryuto Koike, Mengsay Loem, Naoaki Okazaki. Likelihood-based Mitigation of Evaluation Bias in Large Language Models. The 30th Annual Meeting of The Association for Natural Language Processing (NLP2024). 2024. | Young Researcher’s Encouragement Award 🎉 |

Ryuto Koike, Masahiro Kaneko, Naoaki Okazaki. How You Prompt Matters! Even Task-Oriented Constraints in Instructions Affect LLM-Generated Text Detection. The 30th Annual Meeting of The Association for Natural Language Processing (NLP2024). 2024.

Ryuto Koike, Masahiro Kaneko, Naoaki Okazaki. OUTFOX: LLM-generated Essay Detection through In-context Learning with Adversarially Generated Examples. The 18th Symposium of Young Researcher Association for NLP Studies (YANS2023). 2023. | Sponsorship Awards 🎉: PKSHA Technology and HAKUHODO Technologies |

Ryuto Koike, Masafumi Hagiwara. A Support System for Generating YouTube Titles using Style Transfer. The 23rd Annual Meeting of Japan Society of Kansei Engineering. 2021.

🎓 Education

Apr 2023 - Present:                Doctor of Engineering (Computer Science)              Tokyo Institute of Technology

Apr 2021 - Mar 2023:            Master of Engineering (Computer Science)              Keio University

Apr 2017 - Mar 2021:            Bachelor of Engineering (Computer Science)           Keio University

🏆 Honors and Awards

Tokyo Tech SPRING Scholarship Awardee
Tokyo Institute of Technology, 2024-2026
Scholarship: 2,160,000 JPY / APPROX 14,400 USD per year, Research Funds: 300,000 JPY / APPROX 2,000 USD per year,
Full Tuition Exemption.

Sponsorship Awards from PKSHA Technology and HAKUHODO Technologies (1/140=0.7%)
The 18th Symposium of Young Researcher Association for NLP Studies (YANS 2023), 2023

Tokyo Tech Advanced Human Resource Development Fellowship for Doctoral Students
Tokyo Institute of Technology, 2023-2024
Scholarship: 1,800,000 JPY / APPROX 12,000 USD per year, Research Funds: 300,000 JPY / APPROX 2,000 USD per year,
Full Tuition Exemption.

🧑‍🏫 Teaching Experience

Winter 2023: Machine Learning at the Department of CS, Tokyo Institute of Technology.

📸 Media

Oct 2023: Our work was featured in Nikkei🗞.

🖋 Academic Service

Reviewer: ACL