Blaise Delattre
I am a post-doctoral researcher in the team of Prof.
Yang Cao
at Science Tokyo (Tokyo Institute of Science), working on the trustworthiness and provable robustness of
foundation models.
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.
⬇️ Download my resumé
📘 Read my PhD manuscript
News
- Started a new postdoctoral position in Fall 2025 at Science Tokyo.
- Successfully defended my PhD on 20 June 2025 at Université Paris-Dauphine PSL.
- AISTATS 2025 – "Bridging the Theoretical Gap in Randomized Smoothing" accepted.
- ICLR 2025 (oral) – "Accelerated Training through Iterative Gradient Propagation Along the
Residual Path" accepted.
Research Interests
- Certified Robustness and Trustworthy AI
- Lipschitz Networks & Randomized Smoothing
- Stable and Efficient Training of Deep Models
Publications & Preprints
-
Conditional Distribution Quantization in Machine Learning
Blaise Delattre, S Delattre, A Vérine, A Allauzen
arXiv preprint arXiv:2502.07151, 2025
-
Accelerated Training through Iterative Gradient Propagation Along the Residual Path
E Fagnou, P Caillon, Blaise Delattre, A Allauzen
International Conference on Learning Representations (ICLR), 2025
-
Bridging the Theoretical Gap in Randomized Smoothing
Blaise Delattre, P Caillon, E Fagnou, Q Barthélemy, A Allauzen
The 28th International Conference on Artificial Intelligence and Statistics (AISTATS), 2025
-
Chain and Causal Attention for Efficient Entity Tracking
E Fagnou, P Caillon, Blaise Delattre, A Allauzen
Association for Computational Linguistics (ACL), 2024
-
Spectral Norm of Convolutional Layers with Circular and Zero Paddings
Blaise Delattre, Q Barthélemy, A Allauzen
arXiv preprint arXiv:2402.00240, 2024
-
The Lipschitz-Variance-Margin Tradeoff for Enhanced Randomized Smoothing
Blaise Delattre, A Araujo, Q Barthélemy, A Allauzen
International Conference on Learning Representations (ICLR), 2023
-
Efficient Bound of Lipschitz Constant for Convolutional Layers by Gram Iteration
Blaise Delattre, Q Barthélemy, A Araujo, A Allauzen
International Conference on Machine Learning (ICML), 2023
-
A Unified Algebraic Perspective on Lipschitz Neural Networks
A Araujo, A Havens, Blaise Delattre, A Allauzen, B Hu
International Conference on Learning Representations (ICLR), 2023
-
A Dynamical System Perspective for Lipschitz Neural Networks
L Meunier, Blaise Delattre, A Araujo, A Allauzen
International Conference on Machine Learning (ICML), 2022
Contact Information