Koki Wataoka Responsible AI Team, SB Intuitions / Softbank
About
I lead the Responsible AI Team at SB Intuitions, working on AI Safety, especially red-teaming and guardrails for LLMs and VLMs. Previously worked on fairness and privacy‑preserving computation. My recent work studies LLM‑as‑a‑Judge biases (verbosity / self‑preference), jailbreak attacks, and LLM fingerprinting.
Career
2023 – Present
SB Intuitions
Responsible AI Team
2021 – 2023
LINE Corporation
Trustworthy AI Team
2019 – 2021
Kobe University Graduate School
Graduate School of System Informatics
2015 – 2019
Osaka City University
Department of Electrical and Information Engineering
Books
Appointments
Publications
International (peer‑reviewed)
SafeGrad: A Severity-Graded Benchmark for Safety Evaluation, Defense, and Attack of Text-to-Image Models
The Guardrail Trinity: Procedural Safety Schemas, Reasoning Chain Decomposition, and Dual-Mode Inference for LLM Guardrails
Do VLMs Share Safety Neurons Across Modalities?
Domestic (Japan)
公平性が引き起こす敵対的攻撃に対する脆弱性
公平性により生じる敵対的攻撃に対する脆弱性
Lectures / Presentations
June 2026 — Waseda University, School of Fundamental Science and Engineering — Guest Lecturer, "Natural Language Processing" Course
Patents
Contact & Links
Email: wataoka.koki@gmail.com
Google Scholar · GitHub · X · Instagram