📢 Actively looking for full-time researcher positions starting in April 2026.
I am Ryuto, a final-year PhD student (est. March 2026) at the Institute of Science Tokyo, advised by Prof. Naoaki Okazaki. During my PhD, I have engaged in multiple research projects with leading scholars in NLP, including Prof. Chris Callison-Burch at Penn NLP (as a visiting student) and Prof. Preslav Nakov at MBZUAI NLP (remote). In addition to my PhD research, I am involved as a research advisor for a startup on multi-lingual text generation.
I am passionate about making AI systems actually work in the real world, with particular interests in AI safety, Robustness, Interpretability, and Evaluation. My current research focuses on the automated detection of AI‑generated content, specifically on building deployable detectors in practical scenarios with minimum harm. Besides detection, I have also worked on improving the reliability of LLM-as-a-judge in text evaluation:
- 💪 Robustness. Proposed OUTFOX, a detection framework against adversarial attacks [AAAI'24]. Found a vulnerability against distributional shifts via prompt variation in the wild [EMNLP'24 Findings].
- 👀 Interpretability. Proposed an interpretable detector that can present similar examples as evidence [In submission]. Found that NLP experts can detect AI-generated texts with some clues [In submission].
- 📊 Evaluation. Improved LLM-as-a-judge by mitigating self-preference bias [ACL'24 Findings].
- Personal Website: sites.google.com/view/ryutokoike/
- Twitter: @sponddd
- email: my_first_name.my_last_name[at]nlp.c.titech.ac.jp