Author: croyer

Three papers accepted at AISTATS 2025 that will be held in May 2025 in Mai Khao, Thailand:

• Bridging the Theoretical Gap in Randomized Smoothing, by Blaise Delattre, Paul Caillon, Quentin Barthélémy (Foxstream), Erwan Fagnou and Alexandre Allauzen.
• Unveiling the Role of Randomization in Multiclass Adversarial Classification: Insights from Graph Theory, by Lucas Gnecco Heredia, Matteo Sammut, Muni Sreenivas Pydi, Rafael Pinot (Sorbonne Université), Benjamin Negrevergne and Yann Chevaleyre.
• Memorization in Attention-only Transformers, by Léo Dana (ENS Lyon and former MILES intern), Muni Sreenivas Pydi and Yann Chevaleyre.

Congrats to all!

Three papers featuring MILES team members accepted at ICLR 2025, to take place in Singapore at the end of April:

• Accelerated Training through Iterative Gradient Propagation Along the Residual Path, by Erwan Fagnou, Paul Caillon, Blaise Delattre, Paul Caillon and Alexandre Allauzen.
• SCOPE: A self-supervised framework for improving faithfulness in conditional text generation by Song Duong (Sorbonne Université & Criteo AI Lab), Florian Le Bronnec, Alexandre Allauzen, Vincent Guigue (AgroParisTech), Alberto Lumbreras (Criteo AI Lab), Laure Soulier (Sorbonne Université) and Patrick Gallinari (Sorbonne Université & Criteo AI Lab).
• What Makes Large Language Models Reason in (Multi-Turn) Code Generation?, by Kunhao Zheng, Juliette Decugis (Meta AI), Jonas Gehring (Meta AI), Taco Cohen (Meta AI), Benjamin Negrevergne and Gabriel Synnaeve (Meta AI).

Check them out at the conference!

The paper Geometry-Aware visualization of high dimensional Symmetric Positive Definite matrices, co-authored by Thibault de Surrel, Sylvain Chevallier (Université Paris-Saclay), Fabien Lotte (INRIA Bordeaux) and Florian Yger (INSA Rouen), has been accepted for publication in Transactions on Machine Learning Research. Congratulations!

The paper Differentially Private Gradient Flow based on the Sliced Wasserstein Distance, authored by Ilana Sebag, Muni Sreenivas Pydi, Jean-Yves Franceschi (Criteo AI Lab), Alain Rakotomamonjy (Criteo AI Lab), Mike Gartrell (Sigma Nova), Jamal Atif and Alexandre Allauzen, has been accepted in Transactions on Machine Learning Research.

Congratulations!

The MILES team was in Miami (Florida, USA) to present the paper Chain and Causal Attention for Efficient Entity Tracking, co-authored by Erwan Fagnou, Paul Caillon, Blaise Delattre and Alexandre Allauzen, during the 2024 conference on Empirical Methods in Natural Language Processing (EMNLP).

If you missed the conference, make sure to check the paper!

MILES will be at NeurIPS 2024!

The paper Optimal Classification under Performative Distribution Shift co-authored by Edwige Cyffers (former MILES visiting student), Jamal Atif, Muni Sreenivas Pydi and Olivier Cappé (from our neighbors at ENS) was accepted for poster presentation at the conference. Make sure to check it out!

Pierre Wolinski joined the MILES team as a Maître de conférences (associate professor)! Pierre’s research interests revolve around deep learning, with a focus on optimizing the hyperparameters of neural networks. Welcome Pierre!

The paper Exploiting Precision and Recall to assess the quality and diversity of LLMs, by Florian Le Bronnec, Alexandre Vérine, Benjamin Negrevergne, Yann Chevaleyre and Alexandre Allauzen has been accepted to the 62nd Annual Meeting of the Association for Computational Linguistics (ACL 2024)!

This paper adapts the precision and recall metrics used in image generation to the context of text generation, with experiments on large language models (LLMs) such as LLAMA-2 and Mistral. Congrats to all, and especially to Florian Le Bronnec and Alexandre Vérine for leading the effort!

Théo Delemazure, PhD student and MILES member, is first author on two papers accepted at the 33rd International Joint Conference on Artificial Intelligence (IJCAI 2024) :

– Selecting the Most Conflicting Pair of Candidates

– Comparing Ways of Obtaining Candidate Orderings from Approval Ballots

One paper authored by MILES member will be presented at AISTATS 2024, to be held on May 2-4 in Valencia, Spain:

Optimal Budgeted Rejection Sampling for Generative Models by Alexandre Vérine, Muni Sreenivas Pydi, Benjamin Negrevergne and Yann Chevaleyre.