• Assistant Professor at Machine Learning & Systems Laboratory, Graduate School of Information Science and Technology, The University of Osaka.

• Visiting Scientist at RIKEN Center for Advanced Intelligence Project (RIKEN AIP).

• Visiting Scientist at Lattice Lab., Toyota Motor Corporation.

• I received my Ph.D. in Science (Machine Learning) from the University of Tokyo in 2023, under the supervision of Prof. Issei Sato and co-advisory of Prof. Masashi Sugiyama.

About Me

I am a researcher dedicated to enhancing the reliability of machine learning by exploring the theoretical foundations of robustness, generalization, and uncertainty quantification through Bayesian statistics and learning theory.

My research translates these theories into practice. I explore robustness of machine learning, from ensuring provably robust model alignment for LLMs (NeurIPS25) to developing outlier-robust approximate Bayesian computation (AISTATS21). I also investigate the theoretical underpinnings of generalization and calibration, using PAC-Bayes and information theory to analyze calibration error (e.g., NeurIPS24; ICML25) and the role of latent variables in VAEs (NeurIPS25). My work also extends to the foundational analysis of core Bayesian methods, such as scalable variational inference (JMLR21).

Research Keywords

• Provable Robust Methods
  • LLM Alignment (NeurIPS25); Likelihood-free Inference (AISTATS21)
• Generalization Analysis
  • Stochastic Gradient Langevin Dynamics (SGLD) (NeurIPS23); Latent Variable Models (NeurIPS25)
• Calibration
  • Generalization Analysis for Calibration (NeurIPS24, ICML25); Empirical Risk Minimization (Preprint)
• Bias Analysis for Metrics
  • Calibration Error (NeurIPS24, ICML25); Continuous Ranked Probability Score (CRPS) (ProbNum25).
• Bayesian Methods & Others
  • Variational Inference (VI) (JMLR21); Stein Variational Gradient Descent (SVGD) (TMLR25); Physics-informed Neural Networks (PINNs) (NeurIPS24)
• Applications
  • Modeling flight policies of echolocating bats (Preprint); Data Analysis (Sports, Marketing)

CV

My CV is here.

News

• Oct 15, 2025: Our new paper, L2-Regularized Empirical Risk Minimization Guarantees Small Smooth Calibration Error, was made publicly available on arXiv, coauthored by Prof. Futoshi Futami (The University of Osaka / RIKEN AIP / The University of Tokyo).

• Sep 18, 2025: Our two papers, Scalable Valuation of Human Feedback through Provably Robust Model Alignment and Information-theoretic Generalization Analysis for VQ-VAEs: A Role of Latent Variables, have been accepted to NeurIPS 2025! Many thanks to the reviewers for their valuable feedback, and to the ACs, SACs, and PCs for their hard work amidst their busy schedules. And, huge thanks for my collaborators, Masaki, Mike, and Futoshi!

• Aug 12, 2025: Our paper, On the Convergence of SVGD in KL divergence via Approximate gradient flow, has been accepted to TMLR! Many thanks to the reviewers for their valuable feedback, and to the AE for evaluating our work.

• June 18, 2025: Our new paper, Flight trajectory modeling reveals species-specific obstacle avoidance policies in echolocating bats, was made publicly available on bioRxiv, coauthored by Dr. Yu Teshima (Japan Agency for Marine-Earth Science and Technology), Shoko Genda (Doshisha University), Yota Aoki (Doshisha University), Prof. Shizuko Hiryu (Doshisha University), and Prof. Keisuke Fujii (Nagoya University / RIKEN AIP).

• May 26, 2025: Our new paper, Information-theoretic Generalization Analysis for VQ-VAEs: A Role of Latent Variables, was made publicly available on arXiv, coauthored by Prof. Futoshi Futami (The University of Osaka / RIKEN AIP).