Blaise Delattre

Institute of Science Tokyo

Tokyo, Japan

I am a postdoctoral researcher in Prof. Yang Cao’s group at the Institute of Science Tokyo, Japan. My research focuses on trustworthy AI, with an emphasis on certified robustness, randomized smoothing, Lipschitz-controlled neural networks, and robust foundation-model systems.

I completed my PhD at MILES, LAMSADE, Université Paris-Dauphine PSL, under the supervision of Prof. Alexandre Allauzen and Dr. Quentin Barthélemy. My doctoral work focused on Lipschitz-constrained neural networks and certified robustness.

Research interests: Certified Robustness and Trustworthy AI · Lipschitz Networks and Randomized Smoothing · Robust Foundation Models · Stable and Efficient Deep Learning.

⬇️ Download my CV · 📘 Read my PhD manuscript · 💻 GitHub

news

May 11, 2026	ICML 2026 — My paper “Certified Robustness under Heterogeneous Perturbations via Hybrid Randomized Smoothing” has been accepted. See you in Seoul! A second paper I co-authored, “Trading Complexity for Expressivity Through Structured Generalized Linear Token Mixing”, was also accepted — special congrats to the authors and to Erwan Fagnou.
Jan 22, 2026	ICLR 2026 — “Scaling Direct Feedback Learning with Theoretical Guarantees” has been accepted.
Jun 20, 2025	Successfully defended my PhD at Université Paris-Dauphine PSL — “Lipschitz-constrained neural networks and certified robustness” (supervised by Prof. Alexandre Allauzen and Dr. Quentin Barthélemy).
Jan 22, 2025	ICLR 2025 (oral) — “Accelerated Training through Iterative Gradient Propagation Along the Residual Path” has been accepted as an oral presentation.
Jan 21, 2025	AISTATS 2025 — “Bridging the Theoretical Gap in Randomized Smoothing” has been accepted.

selected publications

AISTATS

Bridging the Theoretical Gap in Randomized Smoothing

Blaise Delattre, Paul Caillon, Erwan Fagnou, and 2 more authors

In The 28th International Conference on Artificial Intelligence and Statistics (AISTATS), 2025

Bib HTML

@inproceedings{delattre2025bridging,
  title = {Bridging the Theoretical Gap in Randomized Smoothing},
  author = {Delattre, Blaise and Caillon, Paul and Fagnou, Erwan and Barth{\'e}lemy, Quentin and Allauzen, Alexandre},
  booktitle = {The 28th International Conference on Artificial Intelligence and Statistics (AISTATS)},
  year = {2025},
}

ICLR

The Lipschitz-Variance-Margin Tradeoff for Enhanced Randomized Smoothing

Blaise Delattre, Alexandre Araujo, Quentin Barthélemy, and 1 more author

In International Conference on Learning Representations (ICLR), 2024

Bib HTML

@inproceedings{delattre2024lipschitz,
  title = {The Lipschitz-Variance-Margin Tradeoff for Enhanced Randomized Smoothing},
  author = {Delattre, Blaise and Araujo, Alexandre and Barth{\'e}lemy, Quentin and Allauzen, Alexandre},
  booktitle = {International Conference on Learning Representations (ICLR)},
  year = {2024},
}

ICML

Efficient Bound of Lipschitz Constant for Convolutional Layers by Gram Iteration

Blaise Delattre, Quentin Barthélemy, Alexandre Araujo, and 1 more author

In International Conference on Machine Learning (ICML), 2023

Bib HTML

@inproceedings{delattre2023efficient,
  title = {Efficient Bound of Lipschitz Constant for Convolutional Layers by Gram Iteration},
  author = {Delattre, Blaise and Barth{\'e}lemy, Quentin and Araujo, Alexandre and Allauzen, Alexandre},
  booktitle = {International Conference on Machine Learning (ICML)},
  year = {2023},
}

ICLR

A Unified Algebraic Perspective on Lipschitz Neural Networks

Alexandre Araujo, Aaron Havens, Blaise Delattre, and 2 more authors

In International Conference on Learning Representations (ICLR), 2023

Bib HTML

@inproceedings{araujo2023unified,
  title = {A Unified Algebraic Perspective on Lipschitz Neural Networks},
  author = {Araujo, Alexandre and Havens, Aaron and Delattre, Blaise and Allauzen, Alexandre and Hu, Bin},
  booktitle = {International Conference on Learning Representations (ICLR)},
  year = {2023},
}

ICML

A Dynamical System Perspective for Lipschitz Neural Networks

Laurent Meunier, Blaise Delattre, Alexandre Araujo, and 1 more author

In International Conference on Machine Learning (ICML), 2022

Bib HTML

@inproceedings{meunier2022dynamical,
  title = {A Dynamical System Perspective for Lipschitz Neural Networks},
  author = {Meunier, Laurent and Delattre, Blaise and Araujo, Alexandre and Allauzen, Alexandre},
  booktitle = {International Conference on Machine Learning (ICML)},
  year = {2022},
}